Anda di halaman 1dari 301

Copyright is owned by the Author of the thesis.

Permission is given for


a copy to be downloaded by an individual for the purpose of research and
private study only. The thesis may not be reproduced elsewhere without
the permission of the Author.
PREDICT ION OF FR EE Z I NG AND THAW I NG
TIMES FOR FOODS

A thes i s prese nted in par t i a l fulf i lment of the


requ ir ements for the degree of. Doctor o f
Phi losophy in B iotechno logy a t
Mas sey Un i ver s i ty

DONALD JOHN C LELAND

1 985
ii

ABSTRACT

A study of methods to predi ct the free z i ng and thawing times of both


regu l ar and i rregu l ar shaped foods was made .

Experi ment al thawing data for foods found in the l i t eratur e , w ere
l imi ted in val ue because the e xper i mental cond i t i ons wer e no t
suff i c i ently acc ur atel y meas ured , des cr i bed and controll ed to all ow
m ean i ngful tes t i ng of thaw i ng t i me pred i ct i on methods to be made . A
compr ehens i ve set of 1 82 e xper iment al measur ements of thawi ng time were
made over a w i de range of cond i ti ons usi ng r egul ar shapes made of
Tyl ose , a food anal ogue , and of minced l ean beef . Free z i ng and thawing
exper i ments for irr egu l ar shapes wer e also carr i ed out because of the
pauc i t y of publ i shed experimental data . U s i ng twelve d i f f er ent two­
and thr ee- dimens i onal irregu l ar shaped obj ects 115 e xper i mental
free z i ng and thawi ng r uns were conducted . C omb i n i ng experimental
res ul ts wi th rel i abl e publ i shed e xper im ental dat a f or free z i ng , a dat a
set compri s i ng 5 93 e xperi ments was establ i she d aga i nst whi ch pr edi c t i on
methods wer e t es t ed .

The par t i al d i f f er ent i al equat i ons that mode l the actua l physi cal
process of heat conduct ion dur i ng free z i ng and thawi ng can be sol ved by
the f i ni te d i ff er ence and f i ni te e l ement methods . Tes t i ng of the
f i ni t e el ement method has not been extens i ve , par t i c ul arly f or
thr ee- d imens i onal shapes . Ther efor e a general formul at i on of the
f i ni te el ement method for one- , two- and thr ee-d i mens i onal shapes was
made and impl emen t ed . Both numer i ca l methods accur at el y pr edi ct e d
free z i ng and thawi ng t im es for regular shapes . Suff i c i ently small
spat i al and t i me step i nt er vals cou l d be used so that errors ar i s i ng
f r om the i mpl ement ation of the methods w er e negl i g i bl e compared wi th
exper i men tal and thermal property data uncer t a i nt i es . Gui de l i nes wer e
es tabl i shed to choose s pace and time gri ds i n appl i cation of the f i ni te
el ement method for i rr egu l ar shapes . Adher ence to thes e gu i de l ines
ens ured that pr edi c t i on method error was i nsi gn i f i cant . A s i mpl i f i e d
f i n i t e el ement method was formulat ed and impl emen t ed . ·rt had l ow er
comput ation cos ts but was l ess accurat e than the general formul ation .
Abstract iii

No accur ate , general , but s impl e method for pred i c t i ng thawi ng times
was found i n the l i teratur e . Four poss i bl e appr oaches f or a generally
appl i cabl e , emp i r i cal pr edi ction f o rmul a w er e i nv es ti gated . Each coul d
be used to pred i ct e xper i m ental dat a f or s i mpl e shapes to w i th i n ± 1 1 . 0%
at the 95% l evel of conf i dence . This acc urac y was e qui val ent to that
d i s played by s i m i l ar f ormul ae f or free z i ng t i me pred i ct i on , and was
onl y sl i ghtly i nf e r i or to the accuracy of the best numeri cal metho ds.
A l l four methods ar e recommende d as accurat e pred i ctors .

For mul t i- d imensional shapes ther e w er e two e x i s ti ng geometri c factors


used to mod i fy slab pred i ct i on met ho ds - the e qu i va l ent heat transfer
d imensi onal i ty ( EHTD ) and the mean cond uc t i ng path l ength (MCP) . N ew
emp i r i cal express i ons to ca l culate these f actors for regular shapes
w er e developed that wer e both more accur ate and mor e wi dely appl i cabl e
than the previ ous vers i ons . Pri nc i pl es by wh i ch EHT D and MC P coul d be
determined accur ately for any two- or thr ee-d imensional shapes were
es tabl i she d . The e ff ect of the f i rs t an d second d imens i on wer e
acc ur ately predi cted but lack of s uf f i c i ent data ( due to h i gh dat a
col l ect i on costs ) prevented accur at e mode ll i ng of the e ffect o f the
third d imension for some i rregul ar shapes .
iv

ACKNOWLEDGEMENTS

I woul d l ik e to thank the fol l ow i ng :

- Pr ofessor R . L . Earle , Dr A . C . Cl eland , Dr S . J . Byrne for the ir


superv i s ion and ass i s t ance.

- Andy for his ex tra gu i dance and encourag ement .

- Mr J . T . Alger , Mr P . Shaw, Mr D . W . Coul i ng for the ir sub t l e sk i l l s ,


pat i en t help and exp er i enc e i n bu i l d ing and ma inta in ing equipme n t .

- Mr M . P . F . Loef f en for h i s t i me and ass i s t anc e in compu t i ng mat ters .

- Mr R . Trott for h i s t i me and the P . N . H . B . for the u se of the ir vac uum


mou ld i ng fac i l i t i e s .

- Mr N . B oy d for h i s help i n supply ing tuna for exper imental work .

- M . I . R . I . N . Z . for the resear ch grant that made th i s work poss i b l e and


for u se of the ir compu t i ng fac il i t i es .

- The Un i v ersi ty Grants Commi t tee and New Z eal and Meat Pr oduc ers ' Board
wh ich p rov i ded fi nanc i a l support in the form of scholarsh ips .

- The Mas sey Un i v ers i ty PRIME 750 compu ters for many m i l l ion seconds of
the ir t i me .

- F ami l y and fr i end s for the ir con t i nual support and he lpfulness .

- J oanne for p roof- read i ng , encouragement and h eap s of TLC .


V

TAB LE OF CONTENTS

ABSTRACT • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • ii

ACKNOWLEDGEMENTS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1v

TABLE OF CONTENTS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • v

LIST OF TABLES . • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • xi

L I ST OF F I GURES • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • xvi

INTRODUCTI ON • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1

2 L I TE RATURE REV IEW • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 3


2. 1 PHASE CHANGE IN FOODS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 3
2 .2 P HASE CHANGE FORMULATI ON • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 6
2.2. 1 Govern i ng Part i al D if f erent i al Equat i ons • • • • • • • • • • • • • • • 6
2.2.2 I ni t i al and Surface Boundar y Condi tions • • • • • • • • • • • • • • • • 8
2.2.3 C ompl et i on o f Phase Change • • • • • • • • • • • • • • • • • • • • • • • • • • • • 10
2.2.4 Symmetry Cond i tions • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 11
2.3 SOLUTIONS USING THE ASSUMPTION OF A UNI QUE PHASE
CHAN GE TEMPERATURE • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 12
2. 3 . 1 Exact Sol uti ons • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 13
2.3.2 Approximat e Sol utions for Slabs • • • • • • • • • • • • • • • • • • • • • • • 13
2.3.3 Approximat e Sol uti ons for R ad i al Geomet r y • • • • • • • • • • • • • 15
2.3.4 Approximat e Sol utions for Mul t i-D i mensional Shapes • • • • 15
2.3.5 Emp i r i cal Approaches • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 16
2.3.6 U s e of Anal ogues and Graphi cal M ethods • • • • • • • • • • • • • • • • 17
2.3.7 Numer i ca l Sol ut i ons • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 17
2.4 SOLUTI ONS FOR PHASE CHANGE OVE R A RAN GE OF TEMPE RATURES • • • • • • 20
2. 4. 1 Approximat e Anal yti ca l Soluti ons for Alloy
Sol i d i f i cation • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 21
2.4.2 Sem i -Anal yt i cal Solut i ons • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 21
Tabl e of Cont ents vi

2.�.3 Empi r i cal Solut i ons • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 23


2.�.� N umeri cal Metho ds • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 2�
2.§ THE EFFECT OF GEOMETRY ON FREEZ IN G AND THAWING
TIME PREDICTI ONS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 27
2.6 PHYS ICAL P ARAMETE RS REQU I RED FOR CALCU LATION OF P HASE
CHANGE I N FOODS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 29
2. 6. 1 Thermal Proper t y Data • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 29
2 . 6. 2 Sur f ace Heat Transf er C oeff i ci ents • • • • • • • • • • • • • • • • • • • • 30
2 .7 SUMMARY • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 31

3 RESEARCH OBJECTI VES • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 33

� NUMERI CAL METHOD FORMULAT I ONS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 37


4.1 INT RODUCT I ON • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 37
�. 2 THE F IN ITE ELEMENT METHOD • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 38
�. 2 . 1 F i n i te E l ement Method Formulat i on • • • • • • • • • • • • • • • • • • • • • 39
�. 2. 2 C omput er I mpl ement ation • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • �2
�. 2. 3 F ini te E l ement M ethod Tes t i ng • • • • • • • • • • • • • • • • • • • • • • • • • �5
�.3 T HE F IN I TE DIFFE RENCE METHOD • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 51
�. 3. 1 F i ni t e C yl i nder F i ni te D if f er ence Scheme Formul at i on • • 52
�. 3 . 2 C omput er I mpl ement ation and T es t i ng • • • • • • • • • • • • • • • • • • • 5�

5 EXPERI MENTAL P ROCEDUR E AND DATA COLLECTION • • • • • • • • • • • • • • • • • • • • • • • • 57


5.1 E XPERI MENTAL ERROR • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 57
5.2 CHOI CE OF P HASE CHANGE MATERI A LS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 58
5.3 TEMPERATURE MEASUREMENT AND CONTROL • • • • • • • • • • • • • • • • • • • • • • • • • • 60
5.4 THAWIN G OF SLA BS . • . . . • . . • • • • . • . • . . . • • • • . • . • • • . • • • • • • • • • • • • • • • 65
5. 4 . 1 The Equ i pment . • . . . . . • • . . . • • • . . . . . • . • . . . . . • • . • . • • • • • • • • 65
5 . �. 2 Thi ckness Control and M eas ur ement • • • . • • • • • • • • • • • • • • • • • 66
5 . �. 3 Measur ement and Control of Surface Heat
Transf er C oeff i c i ents • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 66
5.�.� Anal y s i s o f Heat Transfer i n Slabs • • • • • • • • • • • • • • • • • • • • 72
5.5 THAWING OF INFIN ITE CYLINDERS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 78
5.5. 1 The Equi pment • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 78
5.5.2 D i amet er Control and Meas ur emen t • • • • • • • • • • • • • • • • • • • • • • 79
Tabl e of Contents v1 1

5.5.3 Meas ur ement and Control of Sur f ace H eat


Transf er Co eff i ci ents • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • 82
5.5.4 Anal ysi s of H eat T ransf er i n I nf i ni te C yl i nder s • • • • •• • 85
5.6 THAWIN G OF SPHERES• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 87
5 . 6• 1 The Eq u i pm en t . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . . . . . 87
5. 6.2 D i amet er Control and Measur ement • • • • • • • • • • • • • • • • • • • • • • 8 9
5.6.3 Measur ement an d C ontrol of Sur f ace H eat
Transfer Coeff i c i ents• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 8 9
5.6.4 Anal ysi s of H eat Transf er i n Spher es • • • • • • • • • • • • • • • • • • 90
5.7 THAWING OF RECTAN GULAR BRI C KS• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 91
The Equ i pment ••• •••• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 91
D imensional Meas ur ement and Control • • • • • • • • • • • • • • • • • • • 92
M easur ement and C ontrol of Sur f ace H eat
Transf er Coeff i c i ents• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 93
5.7.4 Analys i s of H eat Transf er i n R ectangu l ar Bri cks • •• • • • • 94
5.8 HEAT TRANSFER IN TWO-DI MENSI ONAL I R RE GULAR SHAPES • • • • • • • • • • • • 98
5. 8 . 1 The Equ i pment . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . • . . • • • • 98
5.8.2 D imensional Meas ur ement and Control • • • • • • • • • • • • • • • • • • 1 04
5. 8.3 Meas ur ement and C ontrol of Sur f ace Heat
Tr ansfer Coeff i ci ents • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 0 4
5 . 8. 4 Analysi s of H eat Transf er i n Two-D imens i onal
Irre gular Shapes • • • • • • • • • • • • • • • • • •••• ••• • ••••••••••• • 1 0 6
5.9 HEAT TRANSFE R IN THREE-DI MENSIONA L IRREGU LA R S HA P ES• • • • ••••• 1 0 8
5. 9. 1 The Equ i pment • • •• • • • • • •• • •• • ••• • • • • • • • • • • • • • • • • • • • • • • 1 0 8
5. 9.2 D imens i onal C ont rol and M easur emen t •• • • • • • • • • • • • • • • • • 1 09
5. 9 . 3 Measurement and Control of Surfa ce Heat
Transf er C oeff i c i en t s•••• •••••• • ••••• • • • • • • • • • • • • •• • • 1 1 0
5.9.4 Analysi s of H e at Tr ansf er for Thr ee-D imens ion al
Irregu l ar Sha pes . . . . . . . . • . • • • • . . . . . . . . . . . • • . • . . . . • • . . 1 1 1

6 EXPERIMENTAL DESIGN AND RESULTS• • • • •• • • • • • • • • • •••• • • • • • • • •• • • • • • • 1 21


6. 1 INTRODUCTION••••• • • • ••• • ••• • • • •• • • • • • • • ••• • • • • • • • • • • • • • • • • • • 1 2 1
6.2 THAWING OF SLABS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 2 4
6.3 THAWING OF INFINITE CYLINDERS ••• • • •••• • • • • • • • • • • • • • • • • • • • • • • 1 25
6.4 THAWING OF SPHERES•••• • • • • • • •••• • •••• • • • •••• • • • • • • • • • • • • • •• • 1 25
Tabl e of C onten t s vi i i

6.5 T HAWIN G OF RECTAN GULAR BRICKS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 2 6


6. 6 TWO-DI MENS IONAL IRREGU LAR SHA P ES • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 2 6
6.7 THREE- DI MENSI ONAL I R RE GULAR SHAPES • • • • • • • • • • • • • • • • • • • • • • • • • • 1 27

7 PREDI CTION OF THAWIN G T IMES FOR SLABS , INF IN ITE CYLINDERS


AND SPHERES • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 4 8
1. 1 VERIFI CATION OF A UN IFIED APP ROACH FOR S IMPLE SHA P ES • • • • • • • • 1 4 8
7.2 PREDI CTION BY NUME RI CAL METHODS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 51
7.3 P REDICTI ON B Y SIMPLE FORMU LAE • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 54
7.3.1 Ex i st i ng Pred i ct i on Formul ae • • • • • • • • • • • • • • • • • • • • • • • • • 1 54
7.3.2 Impr oved Pre d i c tion M ethods • • • • • • • • • • • • • • • • • • • • • • • • • • 1 5 8
1. 3 . 3 Compar i son W i th Free z i ng T ime Pred i ct i on Formul ae • • • • 1 62
1. 4 SUMMARY • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 63

8 PREDI CTION OF FREEZING AND THAWIN G TIMES FOR MULT I -DI MENSIONAL
S HAPES BY NUMERI CAL METHODS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 68 ·

8. 1 INT RODUCTION • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 68
8.2 PREDI CTI ONS FOR REGU LAR SHAPES • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 68
8.3 PREDI CT IONS FOR TWO-DIMENS IONAL I RREGU LAR SHA P ES • • • • • • • • • • • • 1 70
8.4 PREDICTI ONS FOR THREE- DIMENSI ONAL I R REGULAR S HAP ES • • • • • • • • • • 1 71
8.5 FIN ITE ELEMENT METHOD USER GUI DELINES • • • • • • • • • • • • • • • • • • • • • • • 1 76
8.6 SUMMARY • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 7 7

9 PREDI CTION OF FREE ZING AND THAWING TIMES FOR MU LTI-DI MENSIONAL
REGULAR SHAPES BY SIMPLE METHODS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 80
9.1 ANALYTICAL TREATMENT OF THE EFFECT OF GEOMETRY • • • • • • • • • • • • • • 1 80
9.2 FEAS IBLE GEOMETRI C FACTORS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 81
9.3 VERI FICATION OF THE EFFECT OF ENVI ROMENTAL CONDITIONS
ON GEOMETRIC FACTORS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 83
9.4 DEVELOPMENT OF IMPROVED GEOMETRI C FACTOR S FOR
MULTI- DIMENSI ONAL REGU LAR SHAP ES • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 84
9.5 TESTING OF IMPROVED GEOMETRI C FACTOR S AGA INST EXPERI MENTAL
DATA FOR MULT-DI MENSI ONAL REGULAR SHAP ES • • • • • • • • • • • • • • • • • • • • 1 89
9.6 TESTING OF IMPROVED GEOMETRIC FACTORS IN COMBINATION WITH
SIMPLE PREDICTION FORMU LAE • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 91

/
Tabl e of C ont ents ix

9.7 SUMMARY • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 92

1 0 PRE DI CTION OF FREE Z ING AND THAWIN G T I MES FOR MU LT I -DI MENSIONAL
I R RE GU LA R SHAP ES BY SIMPLE METHODS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 98
1 0 . 1 INTRODUCT ION • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 98
1 0 . 2 GEOMETRY PARAMETERS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 9 9
1 0 . 3 DEVELOPMENT OF GEOMETRI C FACTOR S FOR MULT I-D I MENS IONAL
I RREGU LAR SHAPES • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 200
1 0 . � TESTING OF GEOMETR I C FACTORS AGAINST E XP ERI MENTA L DATA
FOR MULT I - DI MENSI ONAL SHAPES • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 2 03
1 0 . � . 1 Rectangu l ar Bri ck F reezi ng and Thaw i ng • • • • • • • • • • • • • • • 203
1 0 . � . 2 Free zing and Thawing of Two-D imensi onal
I rr egul ar Shapes • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 203
1 0 . � . 3 Free zing and Thawi ng of T hree-D imension al
I rr egu l ar Sha pes . • . • • . • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 204
1 0 . � . � C ompari son of Slab Predi ction M ethods • • • • • • • • • • • • • • • • 206
1 0 . � . 5 Anal y s i s of Geomet r i c F actor s • • • • • • • • • • • • • • • • • • • • • • • • 206
1 0. 5 SUMMARY • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 207

1 1 TESTIN G OF P REDI CTION METHODS FOR OTHER MATE RI A LS AND DATA SETS • • 2 1 5
1 1 . 1 I NTRODU CT I ON • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 2 1 5
1 1 . 2 NUMERI CAL P REDI CTION METHODS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 21 5
1 1 . 3 SIMPLE P REDICT I ON METHODS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 2 1 7
1 1 • 4 SUMMAR Y • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 2 1 9

1 2 OVE RALL EVALUATI ON OF FREEZING AND THAWIN G T I ME


PREDI CT ION METHODS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 225
1 2 . 1 I NTRODUCT I ON • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 225
1 2 . 2 NUMERI CAL PREDI CTION METHODS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 226
1 2 . 3 SIMPLE P RE DI CT I ON METHODS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 227
1 2 . � OTHER ATTRI BUTES OF S IMP LE P RE DI CTION METHODS • • • • • • • • • • • • • • • 229
1 2 . 5 COMPARISON OF NUMERICAL AND SIMPLE P RE DI CT I ON METHODS • • • • • • • 230
1 2 . 6 NON -CONSTANT ENVIRONMENTAL CONDIT IONS • • • • • • • • • • • • • • • • • • • • • • • 230
1 2. 7 SUMMARY • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 2 31
Tabl e of C ont ent s X

13 CONCLUSIONS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 23-4

NOMENCLATURE • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 236

REFE RENCES • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 240

APP ENDI X A . SUMMARY OF PUBLISHE D SOL UTIONS TO PHASE


CHANGE PROBLEMS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 2 60
A. 1 ABBREV IATIONS USED IN TABLES A . 1 TO A . 8 • • • • • • • • • • • • • • • • • • • • • 260
A.2 ANALYT I CA L SOLUTI ONS TO P HASE CHANGE USING THE ASSU MPTI ON
OF A UNI QU E PHASE CHANGE TEMPERATUR E • • • • • • • • • • • • • • • • • • • • • • • • 261
A.3 SOLUTI ONS FOR P HASE CHANGE OVER A RAN GE OF TEMPERATURES • • • • • 268

APP ENDI X B . ESTIMATION O F E DGE HEAT TRANSFE R IN SLAB


THAWIN G EXPERIMENTS • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 272

APP ENDI X C . GEOMETRI C FACTOR DATA FOR MU LT I -DI MENS IONA L S HA P ES • • 276

APPENDIX D . F IN ITE E LEMENT METHOD COMPUTE R PROGRAM L IST INGS


AND DATA P REPARATION NOTES • • • • • • • • • • • • ( M i cr of i che ) • • 282
D.1 F IN ITE E LEME NT METHOD COMPUTER PROGRAM DESCRI PTI ONS • • • • • • • • • 2 82
D.2 FIN ITE ELEMENT METHOD COMPUTE R P ROGRAM DATA P RE PARATION • • • • • 283
D.3 SYMBOLS USED IN T HE F IN ITE E LEMENT METHOD COMPUTE R PROGRAMS . 2 91
D.� FINITE ELEMENT METHOD COMPUTE R P ROGRAM LISTIN GS • • • • • • • • • • • • • 29�
D . �. 1 The Full Formul ation • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 2 9�
D.�. 2 The S impl if i ed Formul at i on • • • • • • • • • • • • • • • • • • • • • • • • • • • 308
D.5 SAMPLE PROBLEM W ITH COMPUTE R PROGRAM DATA AND OUTPUT F ILES • • 31 5

APP ENDI X E . SAMPLE CALCULATIONS FOR S IMPLE FREEZING AND THAWING


T I ME PREDICT I ON FORMULAE • • • • • • • • • • • • • • ( M i cr of i che ) • • 331
E. 1 THAWING T I ME P REDI CT ION • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 331
E. 2 F REEZING TIME P RE DI CT I ON • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 33�

APP ENDI CES D and E are i nc l uded on mi cr of i che at the end of APP ENDI X c.
xi

L IST OF TABLES

3.1 D esirable A t t r i but es of Free z i ng and Thawing T ime Predi c tion


M et hods 35

3.2 Cond i t i ons R equ i red For D er i vat i on of S impl e Freezi ng and
Thawi ng T ime Pred i c t i on Formul ae 3 6.

3.3 Factors Aff ec t i ng Freezi ng and Thaw i ng T i mes 36

4.1 C ompari son of R es ul t s Fr om the F i n i t e El ement Method Progr ams


W i th N eumann ' s Sol ut i on For Thaw i ng of a S lab Subj ect to the
F i rs t K ind of Boundar y Condi t i on 47

4.2 Compar i son of R es ul ts F rom the F i ni te E l ement Method Programs


W i th a Know n Anal y t i cal Sol ution F or Cool i ng of a Slab
Subj ect to the Second K i nd of Boundar y C ond i t i on 48

4.3 Compari son of R es ul t s Fr om the F i n i t e El ement Method Progr ams


W i th a N umer i ca l Soluti on For C ool i ng of a Sl ab Subj ect to
the Third K i nd of Boundar y Condi tion Including Rad i at ion 49

4. 4 Compar i son o f R es ul ts F rom the F i ni te E l ement Method Progr ams


W i th a Known Anal y t i c a l Sol u t i on For Cool ing of a Cube
Subj ect to the Third K i nd of Boundar y C ond i t i on

4.5 C ompari son of R es ul t s Fr om the F i ni t e C yl i nder F i n i te


D if f er ence Met hod Program W i th a K nown Anal yti ca l Sol ut i on
For C ool ing of a F i n i t e Cyl i nder Su bj e c t t o the Thi r d
K i nd of Boundar y Cond i t i on 56

5.1 Thermal Pro per t y D at a Used I n C al c ul ati ons By N umeri cal M ethods 61

5.2 Thermal Proper t y Dat a U sed In Calcul at i ons By Simpl e Formul ae 62

6.1 Typi cal Condi t i ons i n Food Fr ee z i ng and Thawing Processes 1 28


L i st of T ab l es xii

6.2 Experiment al Data For Thawing of Slabs o f Tyl ose 1 29

6.3 Exper i mental D at a For Thaw i ng o f Inf i ni te Cyl i nders of Tyl os e 1 30

6.4 Experimental Data For Thawi ng of Spher es of Tyl ose 1 31

6.5 Exper i ment a l Data For Thaw i ng of R ec t angu l ar Bri cks of Tyl os e 1 32

6.6 Experimental D at a For Free z i ng and Thawi ng o f Two-D i mensional


Irr egul ar Shapes of Tyl os e 1 34

6. 7 Exper imental Data For Freezing and Thawi ng of Thr e e-D i mensional
Irregu l ar Sha pes of Tyl ose 1 36

6.8 Exper iment al D at a For Free z i ng and Thawi ng of Slabs and


Mul t i -D imensi onal Shapes of M i nced Lean Bee f 1 37

7.1 Summar y of Per c entage D i f f er ences Between Experimental Free z i ng


and Thawi ng T imes For S i mpl e Tyl ose Shapes and Free z i ng and
Thawi ng T im es C al c ul ated By Slab , Inf i ni t e C yl inder and
Spher e Versi ons of the F i ni te D if f er ence Method 1 64

7.2 Summary of Per cent a ge D i f f erences Between Experiment al Thawi ng


T imes For Tyl ose Sl abs , Inf i ni t e C yl i nders and Spheres and
Thawi ng T imes C al c ul ated By S imple Pr edi ction Formul ae 1 65

7.3 Summar y of Per centage D if f er ences Between Exper i ment a l Thawi ng


T i m es For Tylose Slabs , Inf i n i t e C yl i nders and Spher es and
Thawi ng T imes Calculat ed By The Bes t Present Methods 1 67

8. 1 Summar y of P er centa ge D i fferences Between Experiment al Free z i ng


and Thaw i ng T ime For Tyl os e Mul t i -D imen s i onal Shapes and
Free z i ng and Thawing T im es Calcul ated By Numeri cal Methods 1 78

8.2 Summar y of Per cent age D if f er ences Bet ween Exper i mental Free z i ng
and Thawing T imes For T yl ose Thr ee-D imensional Irr egul ar
Shapes and Freezi ng and Thawi ng T imes Calcul at ed By the
F i ni t e El emen t Method 179
L i s t of Tabl es x1 1 i

9.1 The Effect of B i , Ste and Pk on the R a t i o of Freezi ng and Thawi ng


Times For Inf i n i t e R ods to the T imes For the Equ i val ent Slab 1 93

9.2 Constants For Pred i c tion of EHTD and MCP 1 94

9.3 Summar y of Percent age D i f f er ences Bet ween N umer i ca l l y Calcul at ed


Free zing T i m es For F i ni t e Cyl i n ders , Inf i ni t e Rods and
R ectangular Bri cks and Free zi ng T imes Cal cul at ed By
S imple Predi ction Formul ae 1 95

9.4 Summar y of Percent age D if f er ences Between Exper i mental Freezi ng


and Thawing T imes For Tylose Rectangul ar Bri c ks and Free zing
and Thawi ng T imes Cal cul at ed U si ng S i mpl e Geometr i c Factors 1 96

9.5 Summar y o f P er c entage D i f f er en c es Between Experi ment al Free z i ng


and Thawi ng T imes For Tyl ose R ec t angul ar Bri cks and Free z i ng
an d Thawing T imes C al c ul ated By S imple Predi c tion Formulae 1 97

1 0 . 1 Par ameters For Cal cul at i on of The E ffect of Geomet r y


F or Irr egul ar Shap es 20 9

1 0 . 2 Summar y of Per centage D if f erences Between Numer i cally Cal cul at ed


Free z i ng T im es For F i ni t e Cyl inders , Inf i ni t e Rods and
Rectangular Bri cks and Free z i ng T i mes Cal cul ated By
Simple Predi c t ion Formul ae 21 0

1 0 . 3 Summar y of Percentage D if f er en ces Between Exper i mental Freezi ng


and Thawing T i m es For Tyl ose Mul t i-D imensional Shapes and
Free z i ng and Thawi ng T imes Cal cul at ed By S impl e
Predi c tion Formul ae 21 1

1 1 . 1 C om pos i te Dat a Set For Tes t i ng o f Free z i ng and Thawi ng


T ime Pred i c t i on Methods 220

1 1 . 2 Summar y of Per centage D if f er ences Between Exper i mental Thawi ng


T im es and Predi cted Thawi ng T i m es For M i nced Lean Beef 222
L i s t of Tabl es xiv

1 1 . 3 Summar y o f P er cent a ge D i ffer ences Between Experimental Fr eezing


T i mes and Pred i ct ed Free z i ng T imes C al cul at ed By Eq . ( 7 . 7 ) 223

1 1 . 4 Summar y of P er centage D i ffer ences Between Experiment al Free z i ng


T imes For Mul t i -D imen s i onal Sha pes and Freez i ng T imes
Calcul ated By S i mpl e Pred i c t ion Metho ds 224

1 2 . 1 Summar y of the Percentage D if f er ences Between Exper i m en t al


Free z i ng and Thawing T imes Fr om a C omposi t e D at a Set and
F ree z i ng and Thawi ng T imes Calcul at ed By N umer i ca l and
S i mpl e Pred i c t i on Methods 232

1 2 . 2 C om par i son of the Esti mat ed Exper imen t al U ncer t a i nt y Bounds and
the Means and 95% Conf i dence Bounds For the Numeri cal and
S impl e F ree z i ng and Thaw i ng T ime Pred i ct i on Methods 233

A. 1 Exact Anal y ti cal Sol u t i o ns Assuming a Uni que Phase


Change Temper at ur e 261

A.2 Appro x imat e Anal y t i cal Sol u t i ons For Slabs Assuming a U n i que
Phase C hange Temper a t ur e 262

A.3 Appro x imat e Anal y ti cal Sol u t i o ns For Radi al Geometry Assum i ng
a U n i que Phase Change T emperatur e 265

A.4 Appro x imate Anal y t i cal Sol utions For Mul t i-D imensional Shap es
Assumi ng a U n i que Phase Change Tempera t ur e 266

A.5 Emp i ri cal Sol u t i o ns Assumi ng a U ni que Phase Change T emper at ur e 267

A.6 Approxi mat e Anal y t i cal Solut i ons For Alloy Sol i d if i ca t i on 268

A.7 Empi r i cal Sol u t i o ns For Phase Chan ge Over a R ange


of Temperatur es 26 9

A.8 Sol u t i ons U si ng Equi v al ent D i amet ers t o Account F or


Irr egu l ar Shapes 271
L i s t of Tabl es XV

B. 1 R esul ts of the F i ni t e El ement Method S imul at ion o f Ed ge H eat


Transf er D ur i ng Thaw i ng of Slabs 27 4

C.1 R esul ts of F i n i t e D i f f er ence Method C al c ul at ions To Det ermi ne


Geometri c Factors For Mul ti -D imens i onal R egul ar Shapes 276

C.2 R esul ts of F i ni t e E l ement Method Calcul a tions To D et ermi ne


Geomet r i c F actors For Mul t i -D imens i onal Irregul ar Shapes 281

D.1 Par ameter s For The F i ni t e El ement Method Progr ams 2 84

E.1 Res ul ts of the Sampl e C al cu l at i on For S impl e Free zi ng and


Thawi ng T ime Pre di c tion Formulae 3 36
xv i

L IST OF F I GU R ES

2. 1 Ice Fra c t i on of Free zabl e Wat er D at a For a Ty pi cal Foodstuff 4

2.2 Thermal Conduc t i vi ty D at a For a Typ i ca l Foods t uff 4

2.3 Appar ent Volumet ric Spec i f i c Heat C a pa c i t y D at a F or a Typi cal


Foodstuff 5

2.4 Enthal py D ata F or a Ty p i cal Foo dst uff 5

5.1 Thermal C onduc t i vi ty D at a For Tylose ( A ) and M i nced L ean


Beef ( B ) 63

5.2 Apparent Volumetri c Spec i f i c H eat Capac i t y D ata For


Tyl ose ( A , C ) and Mi nced Lean Beef ( B ) 63

5.3 Schemat i c D i agram of the Exper i mental Slab Thaw i ng Equ i pmen t 69

5. 4 Construction o f T es t Slabs 70

5.5 Pred i ct i on o f Sur f ace H eat Transf er Coef f i c i ents For Slab
Thawi ng Exper iments 71

5.6 Typi cal T emperatur e Prof i l es For Thermocoupl es Pos i t i oned A t


O r N e ar the Surface o f a Thawing Slab 77

5.7 Breakp o i nt Anal ysi s to E s t i mat e Thawi ng T imes From Thermocoupl es


Not Posi tioned Exactly at the Thermodynami c Centre 77

5.8 Schemat i c D i agram o f the L i qu i d I mmer s i on T ank

5.9 Schematic D iagr am of the Syst em U sed To Hol d and Osc i l l ate
the I n f i n i t e C yl i nders and Two- D imens i onal I rregu l ar Shapes
i n the L i qu i d I mmer sion Tank 80

5 . 1 0 The Sampl e Osc i l l at or and I nf i ni te Cyl i nder Thawi ng Equ i pment


U sed i n the L i qu i d Immersion Tank 81

,
L i st of F i gur es xv i i

5 . 1 1 Schemati c D i a gr am Showi ng the Arr angement of the Pol yst yr ene Foam
Caps and Thermocoupl es L eads For Inf i ni t e Cyl i nder Exper i ments 81

5 . 1 2 The Sampl e Osc i l l ator an d Sphere Shapes Used i n the L i quid


Immer s i on Tan k 88

5 . 1 3 Ty p i cal R ectangul ar Brick Shapes 96

5 . 1 4 Schemat i c D i agrams of Box C or ner Types 97

5 . 1 5 Cross-sections and F i n i t e E l ement Method Gri ds For the


Two- D imens i onal Irregu l ar Sha pes N umber s One and F i ve 99

5 . 1 6 Cross- section and F i ni t e El ement M ethod Gri d For the


Two-D imens i onal Irregu l ar Shape N umber Two 1 00

5 . 1 7 Cross- sec tion and F i n i t e El ement M ethod G r i d For t he


Two-D imens i onal Irregu l ar Shape N umber Three 1 00

5 . 1 8 Cross-sec t i on and F i n i t e El ement M ethod Gr i d For t he


Two-D imens i onal Irregu l ar Shape N umber Four 101

5 . 1 9 Cr oss-sect ion and F i n i t e El ement Method G r i d For the


Two- D imens i onal Irregu l ar Sha pe N umber S i x 1 01

5 . 20 Cr oss-section and F i n i t e El ement M ethod Gri d For t he


Two-D imens i onal Irregu l ar Shape N umber Seven 1 02

5 . 2 1 Cross- sec tion and F i n i t e E l ement M et hod Gri d For t he


Two- D imensional Irregu l ar Sha pe N umber E i ght 1 02

5 . 2 2 Schemati c D iagr am Showi ng t he Method o f Thermocoupl e Insertion


and Pos i t i on i ng W i th i n the Mul t i - D imens i onal Irregul ar Shapes 1 03

5 . 23 The Sampl e Oscill ator and Two-Dimensional Irregul ar Shape


Free z i ng an d Thaw i ng Equ i pment U sed in t he L i qu i d
Immer sion T ank 1 03
L i st of F i gur es xv1 1 i

5 . 2 4 The Pyr ami d Thr e e-D imens i onal Irregul ar Shape F i n i t e El ement
Method Gri d 114

5 . 2 5 The Spher e Thr ee-D imens ional Irregular Shape F i n i t e El ement


Method Gri ds 115

5 . 26 The Egg Thr ee-D imens i onal Irregul ar Shape F i ni te El ement


Method Gri ds 117

5 . 27 The F i sh Thr e e-D imens ional Irr egul ar Shape F i n i t e E l ement


Method Gri d 118

5 . 28 The Sampl e Osc i l l ator and Thr ee-D imensional Irr egul ar Shapes
Used i n the L i qu i d Immer s i on T an k 119

6. 1 A Ty pi cal Temper atur e/T ime Prof i l e For Thaw i ng of Slabs


of Tyl ose 1 38

6.2 A Ty p i cal Temper at ur e/T ime Prof i l e For Thawing of Inf i n i te


Cyl i nders of Tyl ose 1 38

6.3 A Ty pi cal Temper atur e/Time Prof i l e For Thawing o f Spher es


of Tyl ose 1 39

6.4 A Ty p i cal T emper atur e/Time Prof i l e For Thawing of R ect angul ar
Bri cks of Tyl os e 1 39

6.5 A Typi cal Temperatur e/Time Prof i l e For Free z i ng or Thawing of


t he Tyl ose Two-D imen s i onal Irr egu l ar Shape N umber One 1 40

6.6 A Typi cal Temper atur e/T ime Prof i l e For Free z i ng or Thawing of
t he Tyl ose Two-D imens i onal Irregu l ar Sha pe N umber Two 1 40

6.7 A Ty p i cal Temper atur e/T ime Prof i l e For Free z i ng or Thawing of
the Tyl ose Two-D i mensi onal I rregu l ar Shape N umber Thr e e 1 41

6.8 A Ty p i cal Temper at ur e/T ime Pr of i l e For Free z i ng or Thawing of


t he Tyl ose Two-D imensi onal Irr egu l ar Shape N umber Four 1 41
L i s t of F i gur es xi x

6.9 A Ty p i cal Temper atur e/T ime Prof i l e For Free z i ng or Thawi ng of
the Tyl ose Two-D imen s i onal I rregu l ar Shape N umber F i ve 1 42

6 . 1 0 A Ty p i cal Temper at ur e / T i me Prof i l e For Free z i ng or Thawing of


the Tyl ose Two- D imens i onal I rregu l ar Shape Num ber S i x 1 42

6 . 1 1 A Ty p i cal Temper atur e/T ime Prof il e For Free z i ng o r Thawi ng of


the Tyl ose Two-D imen s i onal I rr egu l ar Shape N umber Seven 1 43

6 . 1 2 A Ty pi cal Temper atur e/T i me Prof i l e For Free z i ng or Thawing of


the Tyl ose Two-D imens i onal I rr egul ar Shape N umber E ight 1 43

6 . 1 3 A Typi cal Temper at ur e/Time Prof i l e For Free z i ng or Thawi ng of


the Tyl os e Three-D imen s i onal I rr egul ar Pyrami d Shape 1 44

6 . 1 4 A Ty p i cal Temper ature/Time Prof i l e For Free z i ng or Thawi ng of


the Tyl ose Thr ee-D imens i onal I rr egul ar Spher e Shape 1 44

6 . 1 5 A Ty p i cal Temper atur e/T ime Prof i l e For Free z i ng or Thawi ng of


the Tyl ose Thr ee-D imens i onal I rregu l ar F i sh Shape 1 45

6 . 1 6 A Typi cal Temper ature/T ime Prof i l e For Free z i ng or Thawing of


the Tyl ose Three-D i men s i onal I rr egul ar Egg Shape 1 46

6 . 1 7 A Ty p i cal Temper at ur e/T ime Prof i l e For Thawi n g of Slabs of


M i nced Lean Beef 1 46

6 . 1 8 A Ty p i cal Temper atur e/T ime Pr of il e F or Thawing of Rectangul ar


Br i cks of M i nced Lean Bee f 1 47

6 . 1 9 A Ty p i cal Temper at ur e/T ime Prof i l e For Free z i ng or Thawi ng of


a M inced Lean Bee f Two-D imensi onal I rr egul ar Sha pe 1 47

B. 1 The F i n i t e El ement Method Gri ds Used To I n v es t i gat e the Effect


of Edge H eat Transfer D ur i ng Thaw i ng of Slabs of Tyl ose 27 5

D. 1 The F i ni t e El ement Gri d F or the Sampl e Probl em Showing Node


Pos i t i on i ng and Number i ng 31 8
INTRODUCT I ON

Low temperatur e i s one of the mos t important and common means of f ood
p reserv a t i on . For New Zealand , whose economy is b ased on p rimary
i ndus tr i es and export of peri shable products to d i s t an t mark e t s , th i s
is espec ially tru e . Freez i ng , co ld storag e and sub se quent thaw i ng are
all op erati ons on the froz en food cha i n in wh i ch los ses of food qual i ty
can b e s i g n i f i can t . I mportant roles for the food eng i ne er are
therefore to des i g n , op erate and control equ i pment that ma i nta i ns
p rodu c t qual i ty wh ilst k eep ing p rocess i ng costs low .

Low temp erature preservat i on can , a t b es t , only ma i nta i n food qual i ty


l ev e ls so there has b een ex tens ive research i nto factors affec t i ng food
qual i ty dur i ng freez i ng , cold s torage and thawi ng . The phys i cal ,
chemical and b iolog i cal chang es that occur dur i ng these p rocesses have
b een re lated to the temperature h istory of the f ood and the p rocess
cond i t i ons u sed . For some food s , data sugges t i ng op t imal se ts of
cond i t i ons and equ i pmen t to m i n im i se damag e to the f ood p roduc t have
b een found . To des ign and op t i m i se equ i p ment f or l ow temperature
p reser v a t i on o f food the f ood eng ineer must b e ab le to p red i c t free z i ng
and thaw i ng rates for these cond i t ions . Both op erat i ng and c ap i tal
cos t s can be reduced i f simple and accurat e methods to do th is are
ava i l ab l e .

I n the p ast , freez i ng has b een more impor tant commerc ially than thaw i ng
as tha w i ng has been p r edom i nately a domes t ic p r ac t i c e . Ex tens i v e
research i nt o the phy si cal aspec ts o f freez i ng has p roduced methods to
pred i c t freez i ng t i mes that are sat i sfac tory f or many p r ac t i cal
s i tua t i ons . However , in general , the se methods ar e e i ther spec if i c to
i nd iv i dual f ood p roducts or app l y only to . s i mp l e r egul ar g eome tr i es .
Many fro z en foodstuffs are irregular i n shap e s o further research i s
warranted t o deve l op , tes t and val i date a gen eral me thod that pred i c ts
freez ing t imes for a wide var i et y of shap es .

R ecently , wi th greater quant i t ies of goods b e i ng fr oz en and gr eater


emphasis on further p rocess i ng of froz en p roducts , thawing has b ecome
an i mportant industr ial process . A s i mpl i s t i c approach is to tre at
I ntroduct i on 2

thawi ng as the reverse of freez i ng, but the val idi ty of th i s app roach
is ques t i onab l e as it has not b een p rov en that accurate thawi ng
p red i c t i ons w i l l be ob ta ined . As for freez ing, much of the research
i nto thaw i ng rate and t i me pred i c t ion has been for ind i v i dual p r oduc ts .
There remains the need for research to f i nd a general method f or
pred i c t ion of thawi ng t i mes . I deal l y , the me thod would b e simi l ar to
those used to p red i c t freez i ng t imes . It may i ncorporate the same
t echn i ques to account for p r oduct g eomet ry . Th i s would al l ow the
ent i r e area of p red i c t i on of phase chang e in f oods to b e consi dered as
a whol e .

I f accurate p red i c t i on o f f reez i ng and thaw i ng under a wi de rang e of


cond i t i ons and geometr i es c an be ach i ev ed th i s w ould en ab l e more
eff i c i ent p roc ess equ ipment des i gn w i t h i n the cold cha in and hence
wou l d reduce cos ts . Therefore the p r esen t research into pred i c t i on of
thawi ng t imes and i nto the effect of p roduct shap e on b oth freez i ng and
thawi ng rates i s c l ear l y neede d .
3

2 LIT ERATUR E REVIEW

Phase change i n foods i s a comp l ex process . Mos t research into me thods


for p re d i c t i on of rates of phase chang e has used simp l i f i ed mod e l
sys tems for analys i s . It is therefore convenient to cons i de r the
l i terature sub-d iv i ded accor d i ng to the type of simpl i f i ca t i on made .
I n i t ial ly the proc ess of phase change in foods is examined and
desc r i b ed and the phase chang e p rob lem formul ated .

2.1 PHASE CHANGE I N FOODS

Food stuffs are a comp lex system of wa ter , so lutes and macromo l ecules
but are of ten cons i dered for eng i neer i ng pu rpo ses as two frac t ions ; an
aqueou s so lu t i on and a so l i d comp onen t . Phase chang e i n food s inv olves
i ntr i cate int erac tions b etween the aqueou s and so l i d frac tions . In
many so- cal l ed ''h igh mo is ture" sol i d foods ( for examp l e mea t ,
veg e tab l es and f i sh ) the wa ter i s bound i n a so li d matr ix and there i s
a l mos t n eg l ig ible migra t ion o f wa ter dur i ng freez i ng or thawi ng . Phase
chang e inv ol v es ma inly the aqueou s part chang i ng from i ce to water or
v i c e v er s a , wi th the sol id b e i ng relat ively iner t . The interac t ions i n
the food cause con t i nuou s freez ing point dep ress i on a s i ce sep arates
from the aqu e ous phase , so l a tent heat i s released or absorb ed ov er a
rang e of temp era tures dur i ng phase chang e ( Rolfe 1 9 68 p . 1 8 4 , D i ck erson
1 977 ) .

The therma l conduc t i vity and spec ifi c hea t cap ac i ty of wa t er vary by a
fac tor of ab out thr ee from tho se of i c e . La tent heat i s rel eas ed or
ab sorbed as water chang es phase , wh ilst the sol i d fra c tion is
ess ent ial l y unchange d . There fore the thermal proper t i es of food are
h ighly dep endan t on the frac t i on of i ce in the food ( Me ll or & Sepp i ng s
1 976 , Heldman 1 982 ) . The l atent heat can b e " lumped" t oge ther wi th the
sensible heat to g iv e .an " ap p arent " ( effec tive ) spec ific hea t cap acity
( Comini & Bonacina 1 97 4 ) . For a typical solid high mois tur e food the
relationship b et ween ice frac t i on , the rmal conduc tivity , app arent
specific heat c apaci ty and en thalpy as a func tion of temperature are
shown in Fig s . 2.1 to 2 . 4 ( Me llor 1 978 ) . The area under the p eak in
the specific heat c apacity c u rv e , excl uding the sensibl e heat
L i t er ature Rev iew 4


8... 0.5
Li:

0�-----r----�--�
-•o -30 -20 -10 0 10 20
Temperaf\.re (•c)

Fig . 2 . 1 Ice Frac t ion of Freez ab l e Wat er Data For a Typ ical Food s tuf f .

U'
0

I 1.5


-
·:;::
+=
0
�c
0
U.

E o.s
ID .
.c

0�-----r----�---.---.
-•o -lO -20 -10 0 10 20
Temperature (•c)

F i g . 2 . 2 Thermal Conduc t i v it y Data For a Ty p ical Foodst u f f .


Literature Review 5

1,000

0'

1...,
:::1
� 100
>.
-
·c:;
CJ
a.
CJ
u
-
CJ

I 10
-�
-
·c:;

a.
Vl

1
--40 -30 -20 -10 0 10 20
Temperature (°C)

Fig. 2.3 Apparent Volumetri c Specific Heat Capacity Data For a Typical

Foodstuff.

400

300
-

l...,


lOO
>-
-8"
.c
-


100

0��---r----�--r---�---r---,
-40 -30 -20 -10 0 10 20
Temperature (•c)

F i g. 2. 4 E nthalpy Dat a For a T ypical Foodst uff.


-
Dat um: H - o.o J m 3 at - 40 .0 °C.
Li ter a ture Review 6

cont ributio n , is the l a tent heat componen t .

Freezing and thawing p rocesse s dif fer c onsiderab ly . In freezing of


food there is a quite dis tinc t initial freezing temperature ( T if ) , at
which ice first starts to form. Abov e this temperature thermal
p ro p er ties are relatively cons tant and b elow this temp erature rapid
chang es oc cur as the ice frac tion incre ases . These cha ng es in
p roperties dec line as l ess liquid wa ter remains , until on ly "bound "
wa ter is left ( Moran 1 930 , Riedel 1 96 1 , Rol fe 1 9 68 p . 1 86 , Comini &
Bonac i na 1 97 4 ) . A t lower temp era tures further ice formation is limi ted
and p r operties are relatively cons tant at the so cal led fr oz en phase
values . I n thawing the temp era ture region where phase chang e oc curs is
entered s l owly and therma l properties chang e gradual l y . As more l atent
hea t is ab sorb ed and more ice me lts , freezing point dep ression is
reduced so T if is approache d . The maj ori ty of l a tent he at is absorbed
at temp eratures j u st b elow T i f giving a fina l rapid chang e in thermal
p rop er ties . This e ffec t is i ndicated by the heig ht and slope of the
apparent spec if ic heat c apac i ty and entha l py curves for ty pical foods
( Fig s . 2 . 3 and 2 . 4 ) .

I n freezing some non-equilib rium b ehaviou r has b een ob served ( Cl eland


et al 1 982 ) .Supercooling of the food below T if p r ior t o in i tial ice
c ry s ta l nuclea tion has p r ob ab l y occ urred . This type of ef fec t i s
unlikel y in thaw i ng as there is n o equivalent nuclea tion p rocess and
the la tent heat is g radual ly absorbed ini tial l y , ov er a larg e
temp er a ture rang e .

2.2 PHASE CHA NGE FORMULAT ION

2.2.1 Governing Partial Dif f erential Equa tions

P hy sica l l y , phase change i n foods i s defined as heat conduc tion in a


solid . This is g ov erned in a thr e e- dimensional volume , V, by the
p a r tial dif ferential equation:

c ::- : [ :: ] :y [ :; ] : [ :: ]
. k + k + . k + Q ( 2 .1 )
L i terature Rev i ew 7

where C • volumetr i c spec if ic hea t cap ac i t y


• f 1 ( T,mater i a l )
T • temp erature ( ° C or K )
t • t i me ( s)
k = therma l conduc t i v i t y (W m-1 -
oc 1 )

f 2 ( T , ma ter i a l )
x, y,z � dis tance in the three ax i a l d ir e c t i ons (m)
Q = internal hea t g enerat ion ( W m- 3 )
f 3 ( T , mater i a l )
Non-l i nea r i t y ar ises from the temp erature var i ab l e therma l p roper t i es
and the comp l ex nature of the boundary cond i tions appl i ed . App rox i mate
methods must there fore be u sed , or simp l i fy i ng assump ti ons made , to
enab le an analy t i cal so l u t ion to be found . One such assump t ion i s tha t
the latent hea t i s transferred at a un ique temperature ( T if ) and
thermal p rop er t i es undergo a step chang e in va lues at th is temp era ture
( Carslaw & J a eg er 1 955 , p . 282 ) . The formulat ion b ecomes that of a
mov i ng boundary ( St e fan) p rob lem and the p osi t ion of the phase chang e

v
front ( boundary) i s determined by E q . ( 2 . 1 ) sub j ec t to :

L
a
at
1
.. k +
[ ay az
aT
ax
1
X
+ ax a z
aT
ay
l y + a x ay
aT
az
lz ] lv�
- k-
[ ay az
aT
ax
l x + ax az
aT
ay
1y + a x ay
ar
az
lz ] lvl (2 .2 )

a t the boondary b etween the so l i d and l i qu i d phases


( J m- 3 )
v
where L .. l at en t heat
l volume of mater i a l i n unfroz en state ( m3 )
l • d irec t i o na l cos i ne to ou tward normal

A l though th is assump t ion of a un ique phase chang e temperature is t rue


in some common phase chang e s i tua t i ons such as i ce forma t i on and
mel t i ng , freez i ng and thawi ng of pure l iqu i ds and p ure metal cas t i ng ,
it is not true for phase chang e in most so l i d , h igh mo isture f oods .
The poss i b i l i ty of der iv i ng s i mp l e p red i c t ion me thods by us i ng th i s
assump tion has b ee n exp l ored ex tensi vely . However the methods der i ve d
c a n o n l y b e appr ox i mate f o r phase change i n food .
L i terature Rev i ew 8

2.2.2 I n i t ia l and Sur fac e Boundary Cond i t ions

I n freez i ng and thaw i ng of foods a var i e ty of ini t ial and sur face
b oundary cond i t i ons can app ly . Mathema t i c ally , Eq. ( 2 . 1 ) is sub j ec t to
the f i rst k i nd of boundary cond i t ion ( p rescr ibed sur fac e temperatur e ) :

T • Ta on surfac e , S 1 , for t>O ( 2 . 3 )


the second k i nd of boundary cond i t ion ( p rescr i b ed surface heat f l ux ) :

[ aT aT aT
k - 1
ax X
+- 1 y +- 1 z
ay az J = 4> on surfac e , S 2 , for t>O (2 . 4 )

the th ird k i nd of boundary cond it ion ( conv ec t i v e and rad iat i v e cool i ng

[
or hea t i ng ) :

k
aT
ax X
aT
- 1 +-
ay
1
y + ] •
� con• � rad on surfac e , s ,, for t>O (2 .5 )

the fou r th k i nd of boundary cond i t i on ( ar b i trary sur face temperature ) :

T .. T w on surface , S� , for t>O (2 .6 )


and the i n i t i a l cond it ion :
T Tin in volume , V • for t =O (2 . 7 )
where T a - amb i ent med ium temp era ture (oc)
4> .. p resc r i b ed hea t f l ux (W m -2 )
.. f � ( t , pos i t i o n )
4>c on = conv e c t i v e hea t f l ux ( W m -2 )

hcon<T ac -T )
4> rad - rad i a t i ve heat f l ux ( W m- 2 )
• hrad ( T ar -T )
h con - convec t i v e hea t transfer coeff i c ient ( W m -2 o c- 1 )
"' f ( t , pos i tion )
5
T ac - convec t i v e amb i ent temp erature (oc)
- f 6 ( t , pos i t i on )
h rad - rad i a t i v e hea t transfer coeff i c i ent (W m -2 oc-1 )
- adT � r +T 2 ) ( T ar +T )
T ar • rad iat i v e amb ient temperature ( °C or K )
• f 7 ( t , positi on )
£ • rad i a t i on emi ss i v i ty
- r,( ma ter i a1 )
Tw • surfac e temp erature (oc)
• f , ( t , pos i t i on )
Tin • i n i t i al temperature (oc)
L i terature Rev iew 9

2.2.2.1 Appl i cab i l i ty of Boundary Cond it ions The f i r s t and fou r th


k i nd of b ou ndary cond i t i ons can b e cons i dered as a spec ial case of the
th ird wh ere h con �� and T ac = T w or T ac sT a . In p rac t i ce b ecause some
sur face r es i s tance to heat transfer always ex i s t s , not
inf i n i tely large so the f i rs t and four th boundary cond i t ions are sel dom
the b est boundary descr i p ti ons p ossi b l e . In mos t food free z i ng or
thawi ng s i tuat ions rad i a t ion is not a maj or means of heat t r ansfer so
it is conv e n i ently g rouped w i th conv ec t i on in the th ird k i nd of
boundary cond i t ion . The second k i nd of boundary cond i t ion infrequent ly
occu rs and i s seldom used even und er h ighly controlled and acc urate ly
measured exper i mental cond i t ions . The th i rd k i nd of b oundary cond i t i on
is the most phy si cally rea l i s t i c boundary cond it i on i n food phase
change p roces ses .

2.2.2.2 I n i t ial Cond i t ions Both uni form and non-un iform initial
temp era ture cond it i ons oc cur in p rac t i c e . To der i v e a simple
analy t i cal ty pe of so lut ion a uni form i n i t ial temperature d i s t r ibu t i on
must u sually b e assumed . Numer i cal solu t i ons can handle e i ther u n i f orm
or non-un i form i n i t ial cond i t ions . Use of a me an bulk temperature
( L oe ffen et al 1 98 1 ) all ows the effect of a non-uni form init ial
temperature d i s t r ibut ion on a freez i ng or thaw i ng t i me to be
app rox imated , i f a simple me thod u si ng a uniform i n it i al temperature is
used for calculat ions .

The release of la tent heat ov er a rang e of temp era ture p resents


p r oblems in def i n i ng the ons e t of thawi ng . The ini t ial thawi ng p o i nt
unl i k e the i n i t i a l freez ing p o i nt , T i f • is not well defi ned for food .
Cho ice of i n i t ial t emp eratures ( T i n ) of l ess than - 1 0 °C throughou t the
ma ter ial to be thawe d , will ensure that the p roc ess i ncludes all
s i g n i f i cant latent hea t transfer f or common food s . F o r freez i ng the
equ i v a l en t r es t ra i n t is that T i n mus t be greater than or equal to T i f •
Because of the aqueous na ture of food and freez i ng poi nt d epres s i o n ,
T i f w i l l alway s b e l ess than or equa l to 0 °C . S ignif icant superheat ing
above T i f for freez i ng or subcoo l i ng b elow -1 0 °C for thaw i ng , is normal
i n commerc ial p rac t i c e b ecause or p rior processing or storag e
cond i t ions s o these cond i t i ons are sel dom l i mi t i ng .
L i terature Rev i ew 10

2.2.3 Comp l e t ion of Phas e Change

To comp l e te the de f i n i t ion of free z i ng and thaw i ng p rocesses an end


p o i n t mus t be def i ned ( Cowe l l 1 97 4 ) . If the assump t io n of a mov i ng
boundary i s mad e the end p oint i s c l early wh en the phase chang e front
reaches the thermodynami c cent re of the obj ec t .

The non-d i s t i nc t phase chang e reg ion for foods p rec ludes th is
def i n i t ion . The b est al ternat ive is to cons i der the proc ess comp l ete
when the thermody nami c centre reaches a certa in tempera ture , or when a
certa i n mass av erage temperature is reached ( Purwada r i a & He ldman
1 98 3 ) . The d i sadvantage of usi ng a mass averag e temp erature end p o i nt
i s that the mean bulk temperature i s d i ff icult to measure or es t i mate
i n p rac t i ce wi thou t ex tens i ve temperature data ( Khatchaturov 1 958 ) . In
cont ras t us i ng a spec i f ied f i nal thermodynamic c entre temperature as
the end p oint means:
the phase chang e oper a t i on i s generally conserv a t i vely desi gne d as
the mass average temperature is l ess l i m i t ing than the thermodynam i c
centre temp erature
- a cho i c e of f i nal centre temp erature can be mad e to app ro x imate any
des i red equ i l ibrat ion temperature
- unl ess the thermodynami c cent re mov es ( Fl emi ng 1 9 70 ) there is no
doub t that phase chang e i s comp l e te thr oughout the mate r i a l when the
centre re aches the endpoint temp erature
th i s means of de term i n i ng the end p o i nt is the mos t common and easy
to use .
For these reasons i t was u sed i n the p rese nt work .

F or freez i ng , a numb er of f i na l centre temperatures have b e en u sed


-5 oc ( Ro l f e 1 9 68 p . 1 97 , Cowell 1 9 74 , James et al 1 97 6 ) , -1 0 °C
( Khatcha tur ov 1 9 58 , I . I . R . 1 97 2 p . 36 , Clel and 1 977 , de M i che l is &
Calvelo 1 9 8 3 ) and -1 8 °C ( Mo l eeratanond et al 1 98 2 , de Mi che l i s &
Cal v e l o 1 98 3 , Hung & Thompson 1 98 3 ) . The p r ocess t i me obv iously
l eng thens a s the f i na l thermodynami c cent re temp erature is l owere d . A
freez i ng t i me pred i c t i on me thod w ould i deal ly t ak e account of these
chang es . Methods w i th th is capab il i t y are ava i lab l e ( Cl eland & Earle
1 98�b ) . The b es t cho ic e of the f inal thermodynami c c entre temperatu re
i s mad e to ensure tha t all s i g n i f i cant l atent heat relea se tak es p lace
L i terature Rev i ew 11

ab ove i t , and that i t is commensurate wi th the sub sequent s torage


t emp er ature . The endp oint temperature of - 1 0 °C was chosen for the
p r es ent work as i t mee ts these r equ i rements and th e bulk of publ i shed
freez i ng data ar e for th is temperature ( Cl e land & Earle 1 9 8 4 a ) .

For thaw i ng , an ob v iou s cho ice for the f i na l cen tre temperature i s 0 °C .
L arge non-equ i l ibr ium e f fects are unl ik e ly to occur so phase chang e
mu st b e complete at th is temp era ture , ev en i f the food b ehaves c l os ely
to pu re water , as T i f is alway s l ess than or equa l to 0 °C . Further , in
mos t thawi ng si tua t i ons thaw ing throughou t the mater i al is demanded
( for examp l e , so the ma ter ial can be d i v i ded or used immed iately i n
further p rocesses ) , yet the lowest mass av erag e temp erature i s desi red
to minimise pos t- thaw i ng qual i ty los ses ( James et al 1 976 ) . A
temp erature of 0 °C meets both these c r i t er i a . A b ov e 0 °C only se ns i b l e
heat effec ts need to be cons idered . I f p r ed i c t ions to o ther f i nal
centre temp eratures ar e requ ire d , then s i m i l ar methods to those
ava i l ab l e to adj ust freez i ng ti me pred i c t ions for d i fferent f inal
temp era tures can p robably b e u sed .

2 . 2 .4 S ymme t ry Cond i t ions

In geometr i c shap es where there is an ax i s of rota t i o nal symme try the


formulat ion can be r estated ax i symmet r i cal l y . Equat ions ( 2 . 1 ) , ( 2 . 4 )

[ ]
and ( 2 . 5 ) b ecome:
aT
ra k- ( 2 .8 )

[
at ay
aT
ra k (2 .9 )
ar

[
and

ra k
aT
ar
lr +
aT
ay
ly ] • ra [ � con + � rad ] (2 . 1 0 )

where r rad ial d i s tance from ax i s of r otat ion ( m)


a • 1 for cy l i nd r i cal co-ord inates
• 2 and terms in y are del e ted for spher i cal
co-ord ina tes .

At all boundar i e s w i th no other boundary cond i t i on de f ined the s ymme try


( no net heat t ransfer) b oundary cond i t io n i s assumed to app ly:
L i terature Rev i ew 12

k
[ aT
-- 1 +
aT
1
y +
ar
lz
] z 0 on surface , S 5 , for t>O ( 2 . 1 1 )
ax X ay az

2.3 SOLUT IONS USING THE ASSUMPT ION OF A UN IQUE PHASE CHANGE
TEMP ERATURE

To attemp t analy tical soluti on of phase chang e p rob l ems i t is u sual to


assume that phase chang e occurs at a un i qu e temperature (T i f ) and that
thermal p r oper t i e s undergo a s tep cha nge at th i s temperatur e , f rom
constant froz en values to cons tant , bu t d ifferent , unfro z en values ( or
v ic e v er sa) . Even w i th th i s mov i ng bounda ry assump t i on the phase
chang e p rob lem is still ma thema t i cally non-l i near and no general
ana ly t ical so lu t i on has b een found .

To fac i l i tate analy s i s further phy si ca l assump tions and l i mi tat ions ar e
neces sary ar i s i ng b o th from the ma thema t i c s and from the phy s i cal
p roper t i e s of food s . Generally the dens i ty d ifference b etween the
phases is ignored and a uni form i n i t i al temperature d i str i but i on is
assumed . Only simple geometr i c shap es ( slab s , infi n it e cy l i nders ,
sphere s , i n f i n i te rods and rect angular b r i c k s ) are cons idered. Also
analy s i s is often rest r i c ted to the s impler f i rst and second boundary
c ond i t ions and the i n i t i a l cond i t ion that the ma t er ial i s at the pha se
chang e temperature ( no sup erhea t i ng for freez ing or subcoo l i ng for
thaw i ng ) . Var i a t ion of boundary cond i t ions wi th t i me i s general ly not
ex ami ned excep t by numer i ca l method s or some form of t ime-averag ing
techn i qu e . A l l of the se assump t ions reduce the rang e of appl i c ab i l i ty
of the resul t i ng so lution. Many of the solu t i o ns are not ap p rop riate
for food freez i ng or thaw i ng proc ess e s , al thoug h they may b e in other
phase chang e si tua t ions such as pure metal cast i ng , or melt i ng and
sol i d i f icat ion of pure sub s tanc es . In add i t i on some ana ly t i cal l y
der i v ed so lu t i ons may require numer i cal eva lua t ion by compu ter s o th ere
are b o th phy s i cal assump t ions and numer i cal t runcat io n or round i ng
errors i n these case s . In con t rast , finite d if ference and fi nit e

element numerical methods require no physical as sumpt ions although t hey


i ntroduce numerical errors . T he lack of physical assumptions make the
lat t er group the pref erre d methods i f numerical solut ion is require d.
L i terature Rev i ew 13

The common assump t ion of phase change at a un i que temperature and a


step chang e i n therma l p rop er t i es l ead s to methods that have b ee n shown
to poor ly p red i c t freez i ng processes in foods ( Cleland 1 977 ) . Bank off

( 1 9 6 4 ) , Muehlbauer and Sunder land ( 1 9 6 5 ) , Bak al and Hay ak awa ( 1 97 3 ) ,


Ock enden and Hodgk i ns ( 1 975 ) , Clel and ( 1 977 ) , Hayakawa ( 1 977 ) , Wi lson
et al ( 1 978 ) and Crank ( 1 9 8 1 ) all rev iew solu tions mak i ng u se of th i s
assump t io n .

2.3.1 E x ac t Solu tions

E x ac t solu tions for phase chang e assumi ng a un ique phase chang e


temp erature are summar ised in Table A. 1 . The semi - in f i n i te s l ab
geome try desc r i p tion and the boundary cond i t i ons u sed i n all of these
so lut ions v i r tual ly never occur in p rac t ical food phase chang e
s i tua t ions , s o these exac t solu tions are of li ttle p rac t i cal value
( C leland 1 977 ) . All o th er so lut ions to phase change problems i nvolve
the u se of mathema t i cal techn iques and ap p rox imat ions as we ll as
s i mpl i fy i ng phys ical assump t ions .

2.3.2 Approx imate Solu ti ons for Slab s

For slab s the app rox ima te analy ti cal solu t ions can be c l assed i nto
thre e group s : heat bal anc e integ ral and variat ional techn i ques ,
p er turbat i on and ser i es solu t i o ns , and other ana ly tical app roaches .
These are summari sed i n Tab l e A . 2 .

2.3.2.1 H e at Balance I nt eg ra l and Va riat iona l Techn iques The hea t


b alanc e integ ral ( integral p rofi l e ) tech nique of Goodman ( 1 9 6 4 ) and the
v a r i a t ional techn i que of B iot ( 1 957 ) reduce the set of par t ia l
d iff er ent i a l equa t i ons de f i n i ng the p roblem i nto a set of simpler
ord inary i n t egro-d i f ferent ial equat ions . In b oth me thods an
app rox ima t i o n to the temp era ture p ro fi l e in each phase i s assumed . The
me thod i s sens i t i ve to cho i c e of the app r op r iate temperature p rofile
( Goodma n 1 961 ) . I t i s d iff i cu l t to p red i c t the accu rac y achiev ed b y a
pa rt ic u la r p rofile ( Langford 1 97 3 ) . B ell ( 1 97 8 ) sub div i des the region
and solv es for eac h section s i mul taneou sly , wh il e Alb i n et al (1 976 )
stu dy six sta ges in the p ha se c ha nge p roc es s, in att empt s t o i nc r ea s e

t h e accu rac y of the method. Most solutions are restrict e d as they


Li terature R ev iew 14

cons i der only the first or second k i nd of boundary con d i t i on and the
semi - inf i n i te rather than the f i ni te s l ab , th ey r equ ire nume r i cal
evaluat i on , or they app ly on ly where the i n i t i a l temp era ture i s equa l
to the phase change temperature . The only solut ions for the th ird k i nd
of boundary cond i t i o n that do not requ ire nu me r i ca l i n t egra t i on are
tho se due to Goodman ( 1 958 ) , for a semi - inf i n i te slab where i n i t ial
sup erhea t i ng or subcoo l i ng ar e ignored , and Hrycak ( 1 963 , 1 9 67 ) who
ex tends th i s to the case wi th superheat i ng or subcool i ng for b oth a
homog eneous and strat i f i ed semi - i n fi n i t e mat e r i a l . Hrycak ap p l i es the
quasi -s teady state as sump t i on ( l i near tempera ture prof i l e b etween the
surface and the internal mov i ng phase chang e b oundary ) and assumes tha t
the movement o f the phase change front and the heat pene t ra t ion fro nt
are p roport ional .

2.3.2.2 P er t urbat ion and Series Solu t i ons The b as i s of th is g roup of


solu t i ons i s to assume a se r i es solu t i on and fi t the terms to the
i n i t ial and b oundary cond i t ions . The more comp l ex boundary cond i t i on
of the th ird k i nd i s ana ly sed b u t numer i cal i nteg ra t i on or computer
calculat ion are requ i red b ecau se these analy t i cal solut ions are ted i ous
to use due to the ir complex i ty . Al terna t i v e ly , a l ess accurat e
solut ion i s ob ta ined when only the f i rst f e w terms i n th e s e r i es are
tak en . To apply these methods to food free z i ng and thaw i ng the
assu mp t ion that in i t ial temperature is equal to th e phase chang e
temp erature mu st b e mad e wh i ch i s a maj o r d isadvan tage .

2.3.2 .3 O ther Analyt i cal Approaches A number of ma thema t i cal


techn i ques such as embedd i ng , i n t eg ral t r ans forma t ions , ana ly t i cal
integ ra t i on and s impl i fy ing assump tions such as assumi ng the so lution
to b e of the form that the d i s tance the phase change front has mov ed i s
p ropor t i onal to the sq uar e root of t ime , hav e b ee n u sed to arr i ve at
s i mp l e formulae or se ts o f ord inary d i fferen t ial equat ions requ ir i ng
nume r i cal i nt eg rat ion . These ar e app rox ima te solu tions to the phase
chang e p r ob lem . Oz i s ik ( 1 9 78 ) descr ibes the formulat ion of the pha se
chang e p roblem as a he at conduc tion p roblem w i th a mov ing heat source
and the u se of Green ' s func t ions to sol v e i t . Some oth er solu t i ons
g iv e n in Tab l e A . 2 use a co-ord ina te transforma t ion due to Landau
( 1 950 ) that i mmob i l i ses the phase change front . Plank ( 1 91 3 ) and
London & Seban (1 943 ) mak e the qua s i -steady state hea t conduc tion
L i terature Rev i ew 15

assump t io n , that sens ible heat c apac i ty i n the phase between the ou ter
surface and the phase chang e front is neg l ig ib l e . The temp eratu re
prof i le i n th i s r eg ion is therefore l inear and a s i mp l e analy t i cal
solu t i on i s ob ta ined for the th ird k ind of b ou ndary cond it i on but w i th
the ma t er ial ini t ia l ly at the phase change temperature . Bo th the se
assump tions w i l l tend to y i e ld low p red i c t i ons of phase chang e t imes .
Rutov ( 1 9 36 ) cons i dered the s ens ible heat c apac i ty of the phase change
mater ial a fter phase chang e and derives a correc t i on factor to account
for i t . Cochran ( 1 955 ) and Kre i th & Rom i e ( 1 955 ) b o th cons i dered the
sens i b l e hea t capac i ty lump ed a t a mid-point in the slab , but were
unab le to arr ive at a s i mp l e solu t ion ex cep t for unduly restr i c t i v e
b oundary cond it ions . Ker n ( 1 97 7 ) and Gl asser & Ker n ( 1 978 ) der i v e
bounds for the so lut ion by cons ider i ng the effec t o f th e quas i -s teady
state assump tion and ap p rox imat ing the i nteg ra t ion , by assumi ng
d i f f erent temp erature prof i l es be twe en the sur face and the mov i ng
b oundary .

2.3.3 App rox ima te Solu tions for Rad ial Geome t ry

Many of the app rox ima te techniques u sed to get solu tions for slab s hav e
b een ex tended to the case of rad ial heat f l ow i n inf ini te cy l inde rs· and
spheres , bu t reta in the same p robl ems when app l i ed to food phase
chang e . These so lut ions are l i s ted in T ab l e A.3. For a boundary
cond i t i on of the th ird k ind w i th no i n i t i a l sup erhea t i ng or subcool i ng
Pl ank ( 1 9 1 3 ) , London & Seban ( 1 9 4 3 ) , Kor om i & H i ra i ( 1 970 ) , Sh ih & Chou
( 1 97 1 ) , Sh ih & Tsay ( 1 97 1 ) , T i en ( 1 9 80 ) , Seen iraj & Bose ( 1 98 2 ) and
H i l l & Kuc era ( 1 98 3 ) have der ived solu t ions . Chuang & Szekely ( 1 97 2 )
solve for fusion of i n f i n i t e cy l i nders w i th the i n i t i a l temperature not
equal to the phase change temp erature and the th ird k ind of boundary
cond i t i o n , bu t the method requ ires nu mer i ca l i nt eg ra t i o n of the
r esul t i ng ord inary d i f f eren t ial equa t i o n .

2.3.4 Appro x i ma te Solu t ions for Mul t i -D i me ns i onal Shapes

Where the geomet ry is c ompl ex so that heat transfer i s in more than on e


d ir ec t io n , dev e l op ment of ev en app r ox i ma t e analy t ical solut ions i s v ery
d i f f i c ul t . Solu tions have b ee n l im i t ed to dea l i ng w i th obj e c ts of
regular geomet r i c conf igurat ions . These sol u t ions are mostly of
L i terature Rev iew 16

l i mi ted p rac t ical i mpor tanc e b ecause of the ini t ial and b oundary
cond i t i ons u sed . Tab l e A . 4 summar ises av a i l ab le method s .

Sol u t i ons o f p rac t i cal i mportance for foods are tho se due to Pl ank
( 1 941 ) , T anak a & N i sh imoto ( 1 95 9 , 1 9 6 0 , 1 9 6 4 ) and Shamsundar ( 1 9 8 2 ) .
All u se the quas i -s teady state assump t ion and assume that the sur face
of the obj ec t is i sotherma l . The gene ral ap p roach is to f i nd geometr i c
fac tors that mod i fy so lu t ions for s l ab s . Plank f i nds app r op r i ate
fac tors for phase chang e in infi n i t e rod and rec tangular b ri ck shap es
subj ec t to the th i rd k i nd of bounda ry cond i t i on and the i n i t ial
temp er ature equa l to th e phase chang e temperature , by assuming that the
mov i ng boundary rema i ns paral lel to the ou ter surfac e . Tanak a &
N i sh imoto ex tend th is analy s i s for con i cal shap es , f i n i t e cy l i nders and
t rap ez o i dal bod i es . Shamsundar employs a s i m i lar approach for th e same
p roblem bu t all ows d i ff eren t con figura t i ons of the phase chang e front
to b e cons i dere d . Us i ng mathema t i cally def i ned geome tr i e s , conduc t i on
shap e fac tors suggest simple analy ti c solu t ions s i mi lar to those of
Pl ank . Hahne & Gr igul l ( 1 975 ) tabu late the conduc t i on shape fac tors
for a w i de rang e of reg ular shap es and assumed i nt er fac e geometr i es .

However for mos t mul t i -d i mens ional shapes the temperature is no t the
sa me ov er the surface at any t i me , and the shap e of the interfac e w i l l
change a s phase change p r ogr ess es, so the accuracy of the der i ved
g eome t r i c f ac tors i s sub j e c t to doub t .

2 . 3 .5 Emp i r i cal Approaches

Th is g roup of solu tions hav e resulted from recog n i t i on that ev en w i th


maj or s i mp l i fy i ng assump t i ons analy t i cal so lu t ion of phase change he at
transfer i s suff i c iently d if f i cu l t tha t alterna t i v e app roaches are
wor thwh i l e . The app roaches are var i ed in natur e , and are descr ibed i n
Tab l e A . 5 .

Charm & S l av in ( 1 9 6 2 ) mod i fy Neumann ' s solution for the f i r s t k ind of


boundary cond i t ion to mak e i t app l i c ab l e to the th ird k ind of boundary
cond i t i o n by add ing a f i c t i t i ou s th ick ne ss to the obj e c t to accou nt for
sur fac e res i s tanc e t o heat transfer. Churchi l l & Gupta ( 1 977 ) de f i ne
an effec ti v e spec i f ic heat c ap ac i t y value so s tandard e x ac t ana ly ti ca l
L i terature Rev i ew 17

solut ions for heat conduc t i on w i thout change o f phase wi l l approx i mate
me l t i ng or so l i d if i ca t ion . Baxter ( 1 96 2 ) , for slab s and i n f i n i t e
cy l i nders and Tao ( 1 968 ) , for s l ab s , i nf i n i t e cyl i nder s and sphere s ,
stud ied free z i ng subj e c t to the th ird k i nd o f boundary cond i t ion b u t
wi th no superhea t i ng or subcool i ng us i ng f i n i te d i f ferenc e s . The i r
results are d isplay ed as reg ress i on equa t ions and chart s . Bax ter ' s
so lut ion i s s im i l ar in form to Pl ank ' s equat ion. Shamsundar &
S c r i v isan ( 1 97 9 ) and Shamsundar ( 1 98 1 ) der i v e a si mi larity ru l e so
numer i cal resul ts can be app l i ed ov er a w i de rang e of cond i t i ons
wi thou t add it ional compu tat ions .

None of these solu tions apply wh ere l a tent hea t is not relea sed or
ab sorb ed at a un ique temperature as is the case for foods .

2.3.6 Use of Ana logues and Graph ical Methods

B e fore the use of v ersa t i l e d ig i tal compu ters b ec ame common and
p ower ful numer i cal me thods were avai l ab l e , ana l ogue and g r aph i cal
solu t i ons for pha se chang e were u sed . Using elec tr i c al analogues
London & Seban ( 1 9 43 ) , Cochran ( 1 955 ) , Krei th & Rom i e ( 1 955 ) and Horvay
( 1 96 0 ) inv est igated sens ible heat eff ec ts neg lec ted by the quasi - s teady
s tate assump t ion us i ng a lumped heat capac i ty me thod . L i eb man ( 1 956 )
and St ephan ( 1 96 9 ) a lso worked wi th e lec tr i cal ana logues for phase
chang e he at t ransf er wh i le Hashemi & Sl i ep cev i ch ( 1 967b ) deve lop ed a
d iffus ion ana logue . Some graph i cal methods for phase chang e are due to
Ede ( 1 9 49 ) , Ke l l e r & Bal lard ( 1 956 ) , He i ss ( 1 958 ) , Longwe l l ( 1 958 ) ,
Sunder land & Grosh ( 1 96 1 ) and Sokulsk i ( 1 97 2 b) .

Graph i cal and ana l ogue me thods for so lv i ng freez i ng and thaw i ng
p rocesses ar e no longer important b ec au se o f the relat i ve ease and
accuracy of us i ng numer ical me thods on d ig i tal compu t ers .

2.3.7 Numer ical Solu t i ons

Math e ma t i ca l app ro x imat ions to Eqs . ( 2 . 1 ) to ( 2 . 1 1 ) can be made u si ng


the f in i te d i f ference me thod ( Du s i nb erre 1 9 45 ) or the f in i te e lement
method ( Tu rner e t al 1 95 6 ) . The th ird k ind of b ou ndary c on d i t ion and
i n i t ial superheat i ng or subcool i ng are e as i ly mode l l e d , though many
L i terature Rev iew 18

so lut ions cons i der only s i mp l er cond i t ions .

A numb er of me thod s to determ i ne the p os i tion of the phase chang e front


have been u sed . Fox ( 1 9 75 ) and Crank ( 1 9 8 1 ) rev iew these me thods and
descr i b e the p rinc iples b eh ind the techn iques . Br i e fly , they can b e
d iv ided into s ix group s .

( a ) N ormal f i n i te d i f ference or finite e l ement app r ox i ma t i ons to


E q . ( 2 . 1 ) are u sed throughout the ob j e c t exc ep t for the s ec t ion
wh ich has j u st reached the phase chang e temp erature . The nodes
represen t i ng these r eg ions are a r t i f i c a l l y he l d at the pha se . change
temp eratu re unt il the calcu lated accumu l a t i on or remov al of hea t i s
equ ivalent to the latent heat . Then the pha se change fron t moves
i nto the ne x t sec tion and normal app rox imat ions u si ng the thermal
p r opert i e s of the new phase are re i nst igate d . Resea rchers who u s ed
the f i n i t e d i ffer ence me thod i nc lude Dusi nb erre ( 1 9 49 ) , Tao ( 1 9 6 7 ) ,
Charm ( 1 9 7 1 ) , Cha rm et al ( 1 9 7 2 ) , Ro shan e t a l ( 1 97 4 ) and Tarna wsk i
( 1 976 ) . Z i enk i ew i cz et al ( 1 97 3 ) and Ro lphe & Bathe ( 1 9 8 2 ) u sed
the f in i te element method . A refinement of th i s me thod where the
p os i t i on of the phase chang e front w i th in ea ch reg ion is p red i c ted
and temperatures near the front are es t i mated by i n terpo l a t i on
g i ves more accurate t emp erature p ro fi l es w i th l ess osc i l l a t ory
b ehav i our ( Murray & Land i s 1 9 59 , Se i der & Chur ch i l l 1 9 65 , Te l l er &

Chu rc h i l l 1 9 6 5 , Padmanabhan & Subba Raj i 1 9 7 5 , Forg ac et a l 1 9 7 9 ,


C i chy et al 1 98 1 ) .

( b ) A co-ord ina t e transformat ion ( Landau 1 9 5 0 ) i s emp loyed to g iv e a


v ar i ab l e space g r i d that f ixes the b oundary ( phase cha nge fron t ) at
one node . The mov ement of the fron t i s calcul ated wh i l e normal
app r ox i mat ions are used in each phase . I tera t i v e so lut ion is o f ten
necessary to f i t all the spec i f i ed cond it ions . Murray & Land is
( 1 9 59 ) , Lotk i n ( 1 9 60 ) , He i tz & Westwa ter ( 1 9 7 0 ) , Kroeger & Ost rach
( 1 97 4 ) , Duda et al ( 1 975 ) , Sa itoh ( 1 978 ) , Hsu .e t al ( 1 98 1 ) ,
Heurtaul t et al ( 1 9 82 ) and Moor e & Bayaz i t og l u ( 1 9 82 ) u sed f in i te
d if ferenc e s wh il e Bonnerot & Jamet ( 1 97 4 ) and Ly nch & O'Neill
( 1 98 1 ) emp l oyed f ini te elements . Th i s me thod g iv es more acc urate
est imate of the front p os i tion than ( a) but l ess ac curate
temperature prof i le s ( Murray & Land i s 1 9 59 ) .
L i t erature Rev i ew 19

( c ) A v ar i ab l e t i me s t epp i ng scheme i s used wi th a f ix ed spac e g r id so


that the phase chang e front j ump s a full nodal d iv i s i on wi th each
t i me s t ep . An i tera t i v e approach is r equ ired to app r ox i mate the
b ou ndary cond i t i ons w i th heat b alanc es and temper ature p ro f i l es
def ined by E q . ( 2 . 2 ) ( Doug las & Gal l ie 1 955 , Goodl i ng & Khader
1 9 7 4 , Good l i ng & Khader 1 975 , Spar r ow et al 1 97 6 , Gup ta & Kumar
1 9 80 , Yu en & Kle inmann 1 980 , Gup t a & Kuma r 1 98 1 ) .

( d ) The enthalpy t ransformat ion of Eyres et a l ( 1 9 4 6 ) is u sed to remov e


the non-l inear i ty due to the mov i ng b oundary so that Eq. ( 2 . 1 )

[ ] [ ] [ ]
b ecomes:
aH = � aT +
a aT +
a
k
aT +
Q (2 . 12 )
k k
at ax ax ay ay az az
where H - en tha lpy

I: c dT

f 1 1 ( T , pos i tion )
N ormal fi n i t e d i f ference ap p ro x imat i ons to E q . ( 2 . 1 2 ) can b e u sed
throughou t th e reg ion ( Pr i ce & Slack 1 95 4 , Bax ter 1 96 2 , Lockwood
1 966 , Shamsundar & Sparrow 1 975 ) . The me thod has been re f i ned by
u si ng interpol a t i on to determine the temp erature d istr ibu tion and
i n ter fac e posi t ion more accurately between spac e grid p o i n ts
( Shamsundar & Spar row 1 9 7 6 , Voller & Cross 1 9 8 1 a , 1 98 1 b , 1 9 8 3 , B e ll
1 982 ) . Ich ikawa & K i kuch i ( 1 979 ) and K i kuch i & Ich ikawa ( 1 979 )
u sed the simi lar method of var iat i onal i ne qua l i t i es .

( e ) The i so therm migra t ion me thod t r ansforms the p roblem from that of
determ i n i ng the dep endent var i ab l e , temp erature , as a f unc t ion of
pos i t ion and t i me to that of determ i n i ng the dependen t var i ab l e ,
p os i t i on , as a func tion of time and temperature ( Crank & Gup ta
1 975 , Talmon & Dav i s 1 98 1 , Talmon et a l 1 98 1 , Tal mon et al 1 98 3 ,
Gup ta & Kumar 1 9 8 4 ) .

( f ) A v ar iety of d ifferen t schemes ar e u sed near the phase chang e fron t


to account for l atent heat effects and the mov i ng b ou n dar y
cond i t ion ( Ehr l i ch 1 95 8 , L az ar i du s 1 970 , Mor r i son 1 97 0 , Goodrich
1 97 8 , Ramak r i shna & Sast r i 1 98 4 ) . Wel l ford & Ayer ( 1 977 ) , Rub i n s ky
& Cravahlo ( 1 98 1 ) and Raymond & Rub i nsky ( 1 98 3 ) u se the f i n i te
Li terature Rev i ew 20

e l emen t me thod and include spec ial e l ements or nodes a t the phase
chang e front tha t approx i mate the latent he at e f fe c t by a heat
source or s i nk d i s tr ibu ted ov er the e l ement . A l l en & Severn ( 1 9 6 2 )
u se a r e lax ation method and model l atent hea t r e l ea se as mov ing
heat g enera t i on .

Excep t for the enthalpy transforma t i on f ew so lutions for food freez i ng


and thaw i ng are based on these me thod s , as pha se chang e i n foods
dep ar ts sub stan t i ally from the b as i c assump tion of a unique phase
chang e temp eratur e . O ther numer ical me thods can approx i mate l atent
hea t relea se or ab sorp tion ov er a rang e of temp er ature more c lose ly and
have been u sed in p re f erenc e .

2.4 SOLUT IONS FOR PHASE CHAN GE OVER A RAN GE O F TEM PE RATURES

The second way that phase chang e in food can b e t ak en into account is
by mak i ng ma thema t i ca l a l l owanc e for phase change ove r a range o f
temp eratures . Phase chang e ov er a rang e of temp era tures is the ac tua l
s i tua t ion occurr i ng in food freez i ng and thawi ng . The mos t ge neral
app roach wou ld tak e account of any var i a t ion in apparent volume t r i c
sp ec i f ic heat c apac i ty , C, and the therma l conduc t iv i ty , k, in
E q . ( 2 . 1 ) me an i ng tha t n o phy si c a l assump tions ab ou t the phase chang e
process need b e made . For foods the apparen t volume t r i c spec i f i c heat
cap ac i ty is de f ined at any temp era ture as :
dH
c - (2 . 1 3 )
dt
S impler semi -ana ly ti c al and emp ir i ca l app roaches have a lso b een tak en .

Metal a l l oy sol id i f icat ion i s a s i mi l a r pha se chang e p roc es s . It is


u sua l l y mode ll ed by a two-phase z one of so l i d and l i quid b et ween two
phase chang e f ronts mov i ng through the mater i a l . This app roach is
c los er t o the t rue phys ical s i tuat ion dur i ng phase cha ng e i n food s ,
than the assump ti on of a un ique phase chang e temp era ture , bu t i s st i l l
only appr ox i mate b ecau se i t r equ ires assu mp t i ons ab out the d i s tr i bu t i on
of latent heat across the phase chang e reg ion.
L i terature Rev i ew 21

2.4.1 Appr ox i mate Ana lyt i cal Solut ions for A l l oy Sol id i f icat ion

Approx i mate analy t ical so lut ions for al l oy sol i d i f i ca t ion are
summar i sed in Tab l e A .6 . In alloy sol i d if i ca t i on it i s normally
assu med that l atent heat is released u n i formly ov er an ex tended
temp erature rang e . Error w i l l oc cur in u si ng the se so lutions for phase
change i n food becau se foods do not have constant laten t he at release
w i th resp ec t to chang e i n temp erature .

Muelbauer et al ( 1 97 3) and Hayakawa & Bakal ( 1 973 ) cons ider the th i rd


k i nd of boundary cond i t ion bu t b oth so lutions , though ana ly ti c al i n
natur e , a re comp l ex and i mp rac t i cal t o us e .

2.4.2 Semi -Analy tical Solu t i ons

Many attemp ts have b ee n made to mod i fy ex i s t i ng ana ly tical solu t i ons to


tak e accoun t of the ir l i mi tat ions and g ive them l ess dependence on
assump t i ons made in the ir theoreti cal dev e lopment . Pl ank ' s (1 91 3 )
equat ion for a f i ni te slab not in i t ial ly sup erhea ted or subcoo l ed is
often u se d as a b ase for the se mod if i ca t i ons b ecause of i t s simple
form , i t s app l i cat i on to a rang e of regular shap es ( Pl ank 1 94 1 ) and
b ecause i t is for the b oundary cond i t ion of the th ird k i nd . P l ank ' s
equa t ion can b e wr i t ten as :

t •
�H
I T a-T i f l
[ p �
h
+ R �
k
2
] (2 . 1 4 )
or i n d imension less form ( Tch igeov 1 958 ) :
p R
Fo - --- + (2 . 1 5 )
Bi Ste S te
where Mi • chang e in enthalpy during the p rocess ( J m- 3 )
Ta - amb i ent temperature ( oc )
Tif - i n i t i a l freez ing temperature ( oc )
P ,R • geome t r i c fac tors dep end ing on obj e c t shap e
D • ful l th i ck ness or d iameter
h • surface heat transfer coe ff i c ient
Fo • F our i er numb er
• at/D 2
81 • Biot numb er
• hD/k
L i teratu re Rev i ew 22

Ste • S te fan numb er


• C 1 T a - T i f i / 6H
a = thermal d iffusi v i ty
z k /C

Var iou s app rox i mate me thod s hav e b een emp loy ed to acc ount for
non-cons tan t thermal p r opert ies , s ens ible heat ef fec ts b o th ab ove and
b elow the phase chang e temp erature rang e and the non- i sothermal latent
he at rel ease or ab sorp t io n . Some f i t experi mental data wh i l e others
attemp t correc t i ons b ased on theor e t i cal g rou nd s . Al though some t imes
der ived for a spec i f i c p roduc t or cond i t ion many of the me thod s c an b e
mad e t o apply t o a w i der rang e o f p robl ems . Ki nder & Lamb ( 1 97 4 ) and
Cleland ( 1 97 9 ) rev iew the re l a t i v e mer i ts of these formu l ae for food
freez ing .

Cowe ll ( 1 9 67 ) , Lotz ( 1 97 4 ) , Mascheroni & Ca lv e lo ( 1 9 82 ) , De Mi che l i s &

Calve lo ( 1 9 8 3 ) and Pham ( 1 98 4 a ) all d iv i de the freez i ng process i n to


sep arate p recoo l i ng , phase chang e and temper i ng p er iod s . For the phase
change per i od vers ions of Plank ' s equat i on are use d , wh i lst for the
other p er iods a var i et y of d i fferent techn iques ar e u sed to ca lcu late
the heat conduc t ion wi thou t cha nge of phase . Pham ( 1 98 4 a ) wr i tes
Plank ' s equa t ion in a form analog ou s to Newton ' s law of cool i ng and
u ses th i s for all three per iods . For p recool i ng and temper i ng , th i s i s
the same as the lump ed capac i ty me thod of Cochran ( 1 95 5 ) and London &

Seban ( 1 9 4 3 ) exc ep t that the posi t ion of the lumped heat wi th in the
obj ec t is determine d by the B i ot nu mb er . For all these me thods therma l
p r op e r t ies are chos en or calculat ed by av erag i ng techn iqu e s to model
more acc urately the true hea t transfer cond it ions during each p e r i od .
For s i mi l ar reasons Mo tt ( 1 96 4 ) , Flem i ng ( 1 967 ) , Sokulsk i ( 1 972a ) ,
A l b i n et al ( 1 9 79 ) and Pham ( 1 9 8 4 a , 1 9 8 4 b ) a ll suggested the u se o f a
we igh ted av erage freez i ng temperature or temperature d i f fer ence in
solu t i ons b ased on a unique phase chang e temp era ture , whe n phase chang e
i s ov er a r a nge of temp er a tures such as i n food freez i ng a nd thawi ng .
Levy ( 1 9 8 3 b , 1 98 3 c , 1 9 8 4 ) also c onsider ed the free z i ng and thawi ng
p roc ess es i n three s tages and def ined effec t i v e heat c apac i t i es a nd
thermal conduc ti v i ti e s for each stag e f or use in solu t i o ns for he a t
conduc t ion w i th ou t phase chang e .
L i t erature Rev iew 23

Nagaok a e t a l ( 1 955 ) , Edd i e & Pearson ( 1 958 ) , Levy ( 1 9 58 ) , Ear l e &

Freeman ( 1 9 6 6 ) , F l em i ng ( 1 9 67 ) , Fraz erhurst et al ( 1 9 7 2 ) , Slat ter &

Jones ( 1 97 4 ) and Mel l or & Sepp i ng s ( 1 976 ) for freez i ng and Wa lk er


( 1 970 ) , Van ichse n i ( 1 97 1 ) and Van i chse n i e t al ( 1 97 2 ) for thawi ng , all
p r opose mod i f i cat ions to P lank ' s equa t ion in the form of mul t i p l i c a t i v e
fac tors . These generally a i m t o account f o r se nsi b l e heat e ffec ts that
Pl ank assumed n egl ig i b l e .

Khatcha turov ( 1 95 8 ) and Golov k i n e t a l ( 1 9 7 4 ) g iv e f ormulae b ase d b oth


on f i t t i ng exp er i mental data and ana ly t ical cons i dera t i ons .
S chwar tzberg e t al ( 1 97 7 ) ex ami ne s temperature p ro fi l es du r i ng phase
change for i nf in i te and zero B i ot numb er s . Mod i f icat i ons at
i ntermed iate values of Biot numb er are u sed to p red i c t phase chang e
ti mes . Sas t ry ( 1 984 ) u sed an en tha lpy trans forma t ion and Goodman ' s
( 1 9 6 4 ) hea t balance integ ral technique to app rox imate and solve the
change in en tha lpy prof i les wi th t i me dur i ng par t ial freez i ng and
thaw i ng in slab s . T e ider ( 1 96 3 ) , de M i che l i s & Calvelo ( 1 982 ) and
Mascheron i et al ( 1 982 ) developed me thods to accoun t for d i f f erent
surface boundary con d i t ions on each face dur ing freez i ng of slab s .

2.4.3 Emp ir ical So lu t ions

The d if f i c ul t y in d e r i v ing a general solu tion to phase chang e tha t has


w i de app l i c ab i l i ty , even wi th only a part ial ana ly t i ca l bas i s , has
meant that many emp ir i ca l app roaches have b een p rop ose d . Some of these
are relat iv e l y general in the i r appl icat i o n , wh i lst o ther s are v ery
s p ec if i c to the food p roduc t and/or phase chang e cond i t i on s to wh i ch
they app l y . These lat ter solut ions may only i nterpolate experi men tal
data .

For freez i ng slab s , infi n it e cyl inde r s , spheres a nd rectangular b ri ck s


to a f i na l temp er a ture of - 1 0 °C Cl eland ( 1 9 77 ) and Cl e land & Earle
( 1 97 6 b , 1 977 a , 1 97 9 a , 1 979 b , 1 98 2 b ) emp i r i cally mod i f i ed the g eometr i c
fac tors i n Plank ' s equa t i on by reg ressi on analy si s o f exp er imental
dat a . By a s i m i l ar analy s i s of slab freez i ng data to a f i nal
temp er a tu re of - 1 8 °C Hung & Thomp son ( 1 98 3 ) found d ifferen t
mod i f icat ions .
L i terature Rev i ew 24

Calve lo ( 1 98 1 ) for thaw i ng of s l ab s , Creed & James ( 1 9 8 1 ) for thaw i ng


of b ee f slab s , Hay ak awa et al ( 1 9 8 3 a , 1 9 8 3 b ) for freez ing of f i n it e
cyl i nder and inf i n i t e r od shapes and Succar & Hay akawa ( 1 984 ) for
free z i ng of slab s g iv e reg ress ion equa t ions calculated from results
p red i c ted by numer i cal me thod s .

P ham ( 1 9 8 4 c ) u se d a s i mi l ar app roach t o Pham ( 1984a) bu t emp i r i ca lly


def i nes the av erage phase cha nge temperature as a func t io n of the f i nal
thermody nami c centre temp erature and the amb i ent temperature .

D i f f erent ial forms of emp ir ically deve loped vers ions of P l ank ' s
equa t ion al low Loeffen et al ( 1 98 1 ) and Cl eland & Earle ( 1 9 8 2 c ) to
cons i der t i me var i ab l e boundary cond i t ions by s i mp l e numer ical me thod s .
A non-u n i form i n i tial temp era ture d is t ribu t ion wa s handled by
calculat ion wi th a me an in i t ial temperatur e .

The emp i r ical solu tions a r e summar i sed i n Tab le A . 7 .

For a range of regular shapes two r ec ent s tud i es ( Clel and & E ar l e
1 9 8 4 a , P ham 1 98 4 c ) have shown the methods t o p redic t freez ing times due
to Pham ( 1 98 3 , 1 98 4 a , 1 98 4 b , 1 98 4 c ) and Cleland & Ear l e ( 1 982 b) to b e
accurate by comp ar ison w i th a larg e comp os i te f r ee z i ng data set .

2.4.4 Numer ical Me thods

D i rec t numer ical app rox imat ion o f Eq. ( 2 . 1 ) is an ap p roach requir ing no
assump t ions ab out the phys ical proc es ses of freez i ng and thaw i ng .
Errors may ar i s e from imp rec ise thermal data , and from nu mer i cal
t ru ncat ion and round i ng error s . A wide rang e o f nume r i ca l me thods have
b een u sed .

Those u s i ng the f i n i t e d iff erence method i nc lude: simple exp l i c i t


scheme s where k and C are comb ined and therma l d i ffusi v it y , a, i s t ak en
as a func t ion of temperature ( Earl e & Earl 1 966 , Cul lwi ck & Earl e 1 97 1 ,
B a i l ey & J ames 1 97 4 , He ldman 1 97 4 a , Heldman & Gorb y 1 97 4 , Cha ttopad ha y
1 975 , James et al 1 977 , James & Creed 1 980 , Creed & James 1 98 1 b,
Heldman 1 9 83 ) , exp l i c i t solu t i o ns where k and C are tak en as separate
func t ions of temperatu r e ( James et al 1 9 79 , James & Bai l ey 1 980,
L i terature Rev i ew 25

Mascheron i & Calvelo 1 9 80 , de Mi chel i s & Calve l o 1 9 8 2 , Ma scheron i &

Calvelo 1 9 8 2 , Mascheroni e t al 1 9 82 ) , exp l i c it d iff er enc e formulae


b as ed on the en thalpy transformat ion ( Eyres et al 1 9 46 , Price & Sl ack
1 95 4 , Albasiny 1 9 5 6 , Lock wood 1 96 6 , Cordell & Webb 1 9 7 2 , J osh i & Tao
1 97 4 , Shamsundar & S parrow 1 975 , 1 9 76 ) , ful ly i mpl i c i t and two t i me
l ev e l imp l i c i t schemes ( Crank & N i cholson 1 9 47 , Albas i ny 1 9 6 0 , Hashemi
& Sl i epcev ich 1 967 a , Flemi ng 1 97 1 c , Ste i nhag en & Myers 1 98 3 , Sue car &

Hay ak awa 1 9 8 4 ) and three t i me l ev e l imp l i c i t so lu tions ( B onae ina &

Com i n i 1 97 1 , Bonac ina et al 1 97 3 , Com i n i & Bonac ina 1 97 4 , C le land &

Earle 1 9 7 7 b , 1 97 9 a ) .

Bonac ina & Comi n i ( 1 9 7 1 ) and Cl e land & Ear le ( 1 98 4 a ) c onsi dered the
d i f feren t f i ni te d i f fer ence sche mes and conc luded that the i mp l ic i t
three t ime l ev e l scheme due to Lees ( 1 9 6 6 ) is super ior in t e rms of
acc uracy . It is a lso stab l e and convergen t . The thre e t i me l evel
scheme p red i c te d freez ing data to with in ab ou t ± 1 0 % for slab s , i n f i n i t e
cy l inders , spheres and rect angular b r i ck shap es ( Clel and e t al 1 982 ) .

Methods for the i n f i n i t e cy l i nder and sphere g eome t r i e s are based on


the scheme p r oposed by Albas i ny ( 1 9 60 ) . Where i mpl i c i t methods are
u sed for two- and three-d imensi onal geometry a number of a lternat i ng
d i r ec t ion i mp l i c i t schemes have been used ( Doug las 1 95 5 , Peaceman &

Rachford 1 95 5 , Doug las & Rachford 1 956 , Br ian 1 96 1 , Fa irweather &

Mi tchel l 1 9 65 , Al lada & Quan 1 966 , Flem i ng 1 97 0 , McKee & Mi tchel l 1 97 1 ,


Bonac ina & Comi n i 1 97 3 , Cl e land & Ear le 1 97 9 b ) . All the ab ov e schemes
are l i mi ted to shap es w i th regular conf igura t io n . F ini te d i f ferenc e
schemes to mod e l i r r egu lar g eome try in two d imensi ons ( Fl eming 1 97 1 a ,
Br i sson-Lop es & Dom i ngos 1 979 , Koskela inen 1 97 9 ) and non-homog eneous
mater ials ( Fl emi ng 1 97 1 b ) have b ee n dev e loped but ar e comp l ex .

The f i n i t e e l emen t method has b ee n wi dely u se d for p rob l ems w i th phase


change ( Comi n i et al 1 97 4 b , Com i n i & Del Gu i d i c e 1 9 76 , Com i n i & L ew i s
1 97 6 , L ew i s & Bass 1 97 6 , R eb ell ato et al 1 97 8 , Fr i v ik & Thorberg son
1 98 1 , F r i v i k & Com i n i 1 982 , Hay akawa et al 1 983 b , Purwada r i a & Heldman
1 98 3 ) . The maj o r advantag es of the finite e leme n t method ov er the
f in i te d if f er ence me thod are that it is abl e to c op e eas i l y w i th
i r regular g eometr i e s and heterog eneou s mater i a ls ( Arce et al 1 98 3 ) .
Comp l e x boundary cond i t ions are i ncorporated si mp l y ( Com in i & Bonac i na
L i t era ture Rev iew 26

1 97 4 ) and h ig her order ( quadrat i c or cub i c rath er than l inea r )


app ro x i ma t ions t o temp erature p rofi l es can b e u sed ( Emery & Carson,
1 97 1 ) . D i sadvantages of the f i ni te e lement me thod are tha t the t i me
step is ne cc essar i l y shor ter for stab i l i ty and non-osc i l la t ion ( My ers
1 978 , Bald 1 98 1 ) and that the comp u ter s torage requ i remen ts and
p rocess i ng t ime are g reater ( Yalamanch il l i & Chu 1 9 7 3 ) than for f i n i t e
d i f ferenc es.

Test i ng of the f i n i t e e l ement me thod has demons trated good ac curacy of


pred i c t ion aga inst experi mental da ta for two-di mens ional problems
cons i der i ng freez ing of roads ( Fr i v ik et al 1 977 ) , freez ing of i c e ( Hsu
& P i z ey 1 98 1 ) , freez i ng of food ( Purwadaria 1 980 ) , thaw i ng of f i sh
( Mi k i e t a l 1 9 78 ) and freez ing of f i sh ( Mi k i e t a l 1 9 8 2 ) .

F i ni te e l emen ts are common ly emp l oyed to d i scre t i se i n spac e , but b oth


the fi n i t e e lement me thod ( Bruch & Z y rolosk i 1 97 4 , Chung 1 9 8 1 ) and
f i ni te d i fferenc e sc hemes hav e been u sed in t i me ( Donea 1 97 4 , Wood &

Lewis 1 975 ) . Us i ng fi n i te elements means h igher order time


approx i mat ions can be used bu t s ignif i cant ly i ncreases th e s iz e o f �e
compu t i ng p rob l em . Where thermal p rop er t i es ar e chang ing rap idly the
l i nea r , c entra l ly balanced , three t i me l evel Lees ' scheme has b een
fou nd to be stab le and conv erg ent ( Comi n i e t a l 1 9 7 4 b ) .

In numer ical me thods the calculated heat flow thr oug h the obj ec t
surfac e shou l d equa l the in ternal entha lpy chang e ov er the wh ole phase
chang e proces s . C lo se ag reemen t ( re ferred t o as a heat balanc e ) mus t
b e ach iev e d for a nu mer i ca l me thod t o b e cons idered b oth accurate and
val i d . To ensure a heat balanc e , a number of d i fferent methods to
ca l c ulate the thermal p rop er t i e s ( espec ially in th e l atent he at p eak
temp erature rang e ) hav e been propo sed ( Comi n i et a l 1 97 4 b , Com i n i & Del
Gu i d ic e 1 9 7 6 , Lemmon 1 9 7 9 ) . Morg an et al ( 1 9 78 ) suggest they may not
g iv e b enef i ts over d irec t evaluat ion. Cleland e t a l ( 1 982 ) , s tudy i ng
numer ical p re d i c ti on o f freez ing , d iscov er e d tha t the b et ter f in i t e
d i f ferenc e and f in i te element me thods are prob ably more l i mi ted by
thermal data unc erta inty than numer i ca l app rox i ma t i ons . Both the
f in i te element me thod and the f in i te d i f f er enc e me thod for cont inuously
var i ab l e therma l p roper t i es can model a u n ique phase chang e t emper ature
by appr op r iate cho i c e of the therma l p r operty func t ions ( Bo nac ina et al
L i terature Rev iew 27

1 9 73 ) .

The b oundary elemen t me thod , wh ich i s s i m i l ar to the finite e leme n t


me thod , has been p r oposed for non-l inear trans i en t he at conduc t ion w i th
var iab le p rop er t i es ( Brebb ia & Walk er 1 9 8 0 ) . I t ha s the advan tage of
reduc i ng the d i mens ional i ty of the p r ob l em and there fore the compu t i ng
requ irements ( Wrob el & Breb b i a 1 98 1 ) , b u t wi l l only b e accurate for
shapes wi th l a rge sur fac e area- to-volume rat ios ( Wr ob e l & Br ebb i a
1 9 79 ) . It ha s not b ee n use d for hea t transf er w i th phase cha ng e as i t
has n o p r oven advantages over the f i n i te element me thod for th i s typ e
of p rob lem .

2.5 THE EFFECT O F GEOMETRY ON FR EE ZING AND THAW ING TIME PREDICTIONS

A par t from some f i n i te d i fferenc e me thod s and fi n i t e element schemes ,


me thods for freez i ng or thaw i ng ti me p red i c t ion are near ly all
restr i c ted to regular g eometr i c con fi gurat ions such as slab s , infin i t e
cyl inder s , spheres , inf ini te rods and rec tangular b r i ck s . Many food
p roduc ts ar e not regular in shap e so methods are requ ired to t ak e
account of i r r egular geome try .

B y assumi ng ( a ) a geome t r i cal descr i p tion for the phase chang e fro n t ,
( b ) that the sur face temperature i s constant b u t s t i l l t i me var i ab l e
ov er the surface and ( c ) tha t the qua s i -s teady state assump tion is
app l icab l e i n the pha se between the surfac e and the phase change fron t ,
P l ank ( 1 9 4 1 ) , Tanak a & N i sh imoto ( 1 9 59 , 1 96 0 , 1 964 ) and recently
Shamsundar ( 1 98 2 ) der ived geome t r i c fac tors to acc ount for the obj ec t
configura t i on for many regular shap es . Where the se cond assump tion i s
t rue and the phase chang e front i s para l l e l t o th e sur fac e , the fac tors
dep end d irec tly on the surface area t i mes th ick ness -to-v olume ( AD/ 2V )
rat i o . Ru tov ( 1 9 36 ) and Mot t ( 1 9 6 4 ) b oth ex tended the i de a to u s i ng
the rat i o o f p roduc t surfac e area t imes th i ck ness- to-volume as an i ndex
of the shape in calculat ions for all geome tr i es . Th i s app roach is
accurate f o r slab s , i n f i n i te cyl i nders and spheres w i th the th ird k ind
o f boundary cond i t ion as the surface i s a l ways i s o thermal . irresp ec t i v e
of Biot number . Cl e land & Ear le ( 1 9 8 2 a , 1 9 82b) show i t i s not acc urate
for o ther shapes when the Biot numb er i s non-z er o . Pham ( 1 9 84b) u s ed
the same g ener a l app roach bu t i nc ludes a correct i on to the Blot numb er
L i t era ture Rev i ew 28

b ased on the "mean heat conduc t i ng path " ( MC P ) fr om the c en tre to the
sur fac e of the obj ec t . A me thod to calculate the mean conduc t i ng p a th
l eng th i s g iven for freez ing of rec tangular b ri ck shap es b ased on fi t
to experi mental data . Sokulsk i ( 1 97 2 a ) u ses the AD/ 2V rat i o b u t only
consi ders a sec t i on of the obj e c t near the thermodynami c cen tre .

Cleland & E ar l e ( 1 982b) def ine EHT D ( equ ivalent heat t ransfer
d imens i onali ty ) as the ra t i o of the freez i ng t i me for a slab of
equ ivalent th i ckness to the freez i ng ti me of the shap e in ques t io n ,
under the same i n i t i a l and boundary cond i t ions . The v a lue o f EHT D i s
calculated from the rat ios of d i mens ions for rectangular b r i ck shapes
or from exp er imental data bu t is, like Pham ' s method , esse n t i ally
equ ivalent to the area t i mes th i ck ness -to-volume ra t io w i th a
correc tion for B i ot nu mb ers g reater than z ero .

Both the EHTD and MCP conc ep ts gav e accurate p red i c t ions of
exp er imental data for freez ing o f rec tangular b ri ck s when app l i ed i n
conj unc t ion wi th an accurate p red i c t ion me thod for freez i ng of the
simple slab shap e . N e ither method ha s a tot ally theore t i ca l b asi s , bu t
b oth cons i der the l i mi t i ng ca se s of Bi �o and B i � m . Goodness of f i t to
exper i mental data was also u sed in the i r der ivat ion . Bo th EHT D and MC P
have b een p ostulated as su i tab l e for irregular shap es , bu t th is has no t
b een v er i f i e d .

A c ommon me thod t o account for g eome t ry i n phase chang e o f i r reg ularly


shaped foods has been to app r ox i mate the shape by the c loses t s i mp l e
shap e ( slab , inf i n i te cyl i nder , sphere) and define an equ iv alent
d ia me ter or th i cknes s . Th is me thod is of l imi ted u se as the
app rop r i ate correla t i o n for equ iva len t d iame ter is spec i f i c for each
produc t ty pe and r equ ires experi mental dat a to determine i t accurate l y .
Often the equ ivalent d iame t er is chosen to i nc lude correc tion for
errors i n the p red i c t io n me thod u se d , as well as th e effect of shape on
hea t transfer . A g enera l model to ac count for g eometr i c effec ts w i th in
a calculat i on me thod i s super i or to an experi mental corre l a t io n for a
p ar t i cu lar p roduc t , as i t i s not restr i c ted by p roduc t typ e . Tab l e A . 8
l is ts r e sear cher s who u se equ ivalen t d ia me ters for irregular shap e s .
L i terature Rev iew 29

For heat conduc t ion wi thou t phase change , Smi th e t a l ( 1 96 7 ) , Sm i th et


al ( 1 9 6 8 ) , Clary et al ( 1 9 6 8 ) , Smi th & N e lson ( 1 9 6 9 ) and Cl ary et al
( 1 97 1 ) def ined a geome t ry i ndex for i r r egular shapes calcula ted from an
ell i p so i dal model shap e tha t has equa l orthog ona l cross-sec t iona l ar eas
to the anomalous shap e tha t i t r ep lac es. Th i s i ndex has not b een
tested for phase chang e p rob lems .

2 .6 PHYS ICAL PARAMETERS REQUIRED FOR CALCULAT ION OF PHASE CHANGE


IN FOODS

2 .6 . 1 Therma l Proper ty Data

The therma l p rop er t i es requ ired to p red i c t phase chang e in food i nc lude
thermal conduc t iv i ty , k , volume t r i c spec i f i c heat c apac i ty , C, latent
hea t content , L, phase chang e temp erature , T if ' and spec i f i c hea t
g enera t i on , Q . N o me thod to p re d ic t freez ing or thaw i ng times for
food s tuffs can be accura te i f the food thermal propert i es requ i r ed by
the me thod cannot b e ob tained ac cura tely , or are not rep rese nta ti ve of
the who l e produc t . There has been a great deal o f r esear ch , rev i ewed
by Woolr i ch ( 1 9 6 6 ) , Woodams & Nowry ( 1 96 8 ) , Mor l ey ( 1 97 2 ) , Me llor
( 1 976 , 1 9 78 , 1 979 , 1 980 ) , Pol l ey e t al ( 1 9 80 ) and Me ffer t ( 1 98 4 ) , i nto
thermal p roperty data requ ired to calculate phase chang e for foods . A
comp l ied set of da ta for meat ( Mor l ey 1 972 ) demons t r a tes some
i ncons i stenc ies in the thermal p rop er ty data . S i mi lar i ncons i s tenc ies
in data c ould b e exp ec ted for other food s .

Collec tion of thermal p roperty data i s ted iou s and exp ens i ve b ecau se o f
the v ar i ab i l i ty o f foods tuffs i n compos i t ion and s t ruc ture and the w i de
rang e of temp erature for wh ich data is requ ired . Data for froz en food
are esp ec ially d i f f icult to determine b ecau se it is d i ff icul t to
ma inta in the app ro p riate , exp erimental cond i t i ons .

B ecau se of the se d iff icul t i e s , numer ou s a t temp ts hav e been ma de to


mathe ma t i cally model food p ro p er t i es and hence der i v e calculat ion
me thods for the thermal p r op er t i es . Data that can be determ ined
simply , cheap ly and accurately such as i n it i a l freez i ng temperature ,
compos i t io n , average unfrozen thermal conduc t iv i ty and spec i f ic heat
cap ac i t y and the b ound water frac t i on are u sed as i npu t data f or these
L i t era ture Rev i ew 30

mode ls . The s implest models sum the e ffec t of the sol i d and l iquid
componen t p a r t s ( Com i n i e t a l 1 9 74 a , Lamb 1 9 76 , D i ckerson 1 9 77, L e ntz &
Van Den Berg 1 9 77, Mascheroni e t a l 1 9 77, Levy 1 9 8 1 , Levy 1 9 8 2 a ) . An
al terna t i v e i s to pred i c t the water and ice frac t io n from freez i ng
point dep ress i on equa t i o ns and re late th is to the chang es i n p ro p er t i es
from unfroz en values ( He ldman 1 9 74 b , Heldman & Gorby 1 9 75 , Schwa r tzberg
1 9 76 , Schwar tzb erg 1 9 77, Mascheron i & Ca lv elo 1 9 80 , Heldman 1 9 8 1 ,
Heldman 1 982 , Mascher on i & Ca lvelo 1 98 2 , Lark in e t a l 1 984 ) . A th i rd
app roach i s to u se emp ir i cal or reg ress i on equa t i ons to f i t the data
( Levy 1 9 79 , Chang & Tao 1 98 1 , Ramaswamy & Tung 1 98 1 , Levy 1 9 82 b , 1 982 c ,
1 983a) . All three app roaches are c l a imed to l ead to data of s i mi lar
ac curacy to that ob ta inabl e by d irec t measuremen t .

Numer ical me thod s b ased on approx imat ing Eq. (2 . 1 ) ne ed ex tens i v e


therma l data over the ful l tempera ture range wher eas many s i mp l e
me thods d o no t . Complete therma l p roper ty tab l es , simp l i f ied l i ne a r
app r ox i mat ions t o the ac tual data ( Bonac ina e t al 1 9 74 , Com i n i e t al
1 9 74 a , B onac ina & Comi n i 1 9 76 , Tao 1 9 75 ) and sy stems b ased on the ab ov e
p r op e r ty p red ic t ion me thod s have a l l b een used in numer ical me thods
wi th good rep orted acc uracy .

Meff'e rt ( 1 9 8 4 ) and Hs i eh et a l ( 1 9 77) inv est igated the error i n therma l


p r op e r ty data and i ts effec t on p red i c t ion accuracy . Cons i der i ng tha t
error o n i n d iv i dual values of any p roper t y ca n b e as h igh as 4 0 % the
error can b e substantial in any sub s equen t pred i c t ions of freez i ng and
thawi ng .

2.6 .2 Surface Hea t Transfer Coeffic ients

One of the mos t d iff icult parameters to de termine indep enden tly in food
freez i ng and thaw i ng proc es ses that i nvolve the th ird k i nd of boundary
cond i t ion i s the surface heat transfer coe ff i c ient b et ween the amb i ent
med ium and the p roduc t surfac e . Arce & Sweat ( 1 980 ) r ev iew tech n i ques
to est imate , and g iv e data for , surface hea t transf er coeff i c i e n ts in
many food heat t r ansfer s i tu a t i ons.

There are many correla t i o ns ( McAdams 1 95 4 ) b u t these tend to b e med ium,


p r oduc t , and s i tuat i on spec if i c , and of doub t ful accuracy . Measurement
L i terature Rev i ew 31

i n s i tu i s a b e t te r p r op osa l a s the ac tual heat transfer cond i t io ns a r e


more c lose ly rep rese n ted and any var i a t i on due to equ ipme nt will be
t ak en into ac c oun t . Some t rans i ent methods are d i scussed b y Cleland &

Ear le ( 1 9 7 6 a) and Earle & Cl e land ( 1 97 9 ) . Some research es hav e u sed


metal t ransduc ers ( F lem i ng 1 967 , Kopel man et al 1 97 0 , Ear l e 1 97 1 ,
Cowe l l & Namor 1 9 7 4 ) . Users o f such dev i c es canno t be sure of the
corre l a t ion b etween the sur face heat transfer coeff ic i en t measur ed for
the transduc er and those for the ac tua l foodstuff or mater i a l . Others
work ers hav e used the ac tual food under s i m i lar experi mental cond i t i ons
and hav e chosen the surface heat transf er coef f i c ient to fi t so lu tions
for he at transfer wi th ou t pha se change ( Charm 1 963 , Bak er & Charm
1 969 ) , or to g iv e the b est fi t of surface and cent re temp erature
p red i c t ions us i ng f i ni te d i f ference me thod s ( Beck 1 969 , Bonac i na &

Comi n i 1 97 2 , Chav arr i a & He ldman 1 984 ) . Com i n i ( 1 97 2 ) shows tha t


sur face tempera tures a r e mos t sens i t i v e to change in the sur fac e heat
transfer coe ff i c ient and hence advocates the ir u se in an op timal
exper i mental des ig n . For measur i ng sur fac e heat t ransfer coeff ic i en ts
to irregular shapes on ly me tal tr ansducers are can b e u sed wi thou t the
need for comp l ex tr ial and err or compu ter ana ly s i s by the f in i te
el ement method . F or s i m i l a r rea sons i n s i tu tr ans i e n t de termi na t i ons
are l i mi ted to regular geome tr i c shapes . Var i a t ion of th e surface heat
transfer coeff i c ient ov er the sur face is c ommon w i th var i ab l e g eome t r i c
conf igurat ions and i n cond i t ions such as a ir f low ( P urwada r i a 1 980 ) .
Local v a r i a t i ons of the surface heat transf er coe ff i c ient can b e
s ig n i f i can t b u t are v ery d i ff icul t t o mea sure . Often a n av erage value
mu st b e determ i ned or est i ma t e d .

2 .7 SUMMARY

There ar e a larg e nu mb er of solu tions to the p rob l em of he at con duc t ion


wi th phase change for regu lar geome t r ic shapes . Mo st analy t ical and
some nume r i c al solu t i o ns ar e b ased on the assump tion that phase chang e
occurs a t a un ique temperatur e . None of the se have b e en p rov en
accu ra te for food whe r e phase chang e is ov er a rang e of temp eratures .
O ther common l i m i tat ions of me thods are that th ey do not cons i de r the
mos t common and u se ful th ird k ind of b ou ndary cond i t ion , do not accou n t
f o r i n i t ial temperature not be i ng at the phase chang e temp er a tur e , and
are der ived for semi - i n f i n i t e rather than f in i t e g eometr i c shap es .
Li terature Rev iew 32

Mod if ied analy t i c a l and emp ir ical so lut ions retain �e s i mp l e


analy t ical form , have less deta i led thermal data r equ i re men ts than more
r i g orou s method s , bu t tend to be p roduc t or si tua t ion spec i f i c . Wi th in
th i s group , those formulae of a general na ture hav e been tes ted aga i ns t
exp er imental freez ing data and some hav e shown ac cep tab l e ac curacy .
There has b e en no ex tens i v e tes t i ng of si mp l e p red i c t ion me thods for
thawing .

The f i n i te d i ff erence numer ical me thod is of p rov en accuracy for


freez i ng of r egu l ar sh ap es . The f i ni te e l emen t me thod has b e en ·
formulated and shown to b e super ior to the f i n i t e d if fer enc e me thod for
p red i c t i ng phase change in two-d imens ional irregul ar shap e s . Ne i ther
numer ical method has b een tes ted ex tens i v e ly for thawi ng of food i n
regular shap es or for freez i ng or thawi ng of food s of irregu l ar
g eometry due to the lack of su itab le exp er iment a l data .

S i mp l e analy t ical me thods to al low ca lculat ions for irreg u l ar shapes


hav e b een p rop ose d , but no s i g n i f ican t test ing or r e f i n emen t ag a inst
exper i menta l or numer i ca l l y ca lculated data ov er a rang e o f cond i t io ns
has b een carr i ed ou t .
33

3 RESEARC H O BJECTIVES

U l t imately , v er i f i ca t i on of the acc uracy of any method to p red i c t


freez i ng and thawi ng t i mes mus t · b e made by compar i son wi th accu rate
exp er imental freez i ng and thaw i ng data ( H el dman 1 9 83 , Cl e land & Ear l e
1 98 4 a ) . The mos t useful data s e t w ould b e l a rge and d i ver se i n order
to d i fferen t i ate b et ween error due to imp rec ise k nowl edg e and contro l
of the phase change cond i t ions ( exper i mental error ) , error due to
unce rta inty in thermal p rop er ty data ( data error ) and i naccu racy
ar i s i ng from as sump t i ons or app rox i mat ions made in the der i va t i on of
the p re d i c t i o n me thod ( p re d i c tion error ) .

For freez i ng of regular shapes such as s l ab s , cyl inder s, spheres and


rec ta ngular b ri ck s L i ng e ( 1 97 3 ) , Cl e land ( 1 977 ) , de M i che l i s & Calvelo
( 1 983 ) , Hayakawa et al ( 1 98 3 a ) , Hung & Thompson ( 1 983 ) and Succar &

Hayakawa ( 1 9 8 4 ) hav e publ i shed maj or data sets for a variety of


foods tuffs and food anal ogu es, froz en under a range of cond i t i ons .
Cl eland & Earle ( 1 9 8 4 a ) and Pham ( 1 9 8 4 c ) used an ama lgamated data se t
from the se sourc es to tes t freez i ng t i me pred i c t i on me thods for regular
shap es . This study showed that accurate freez ing t ime p re d i c t i on
me thods ex i s t ed for these shapes .

Data g i ven for i rregular shap ed obj ec ts tend to be p roduc t spec if i c .


In v i r tually all cases pub l i shed in the l i teratur e , deta i l ed
g eometr ical descr i p ti ons have not b een g iven and cons i s t ency in size
and shape dur i ng exper i ments was not control led . For examp l e , i n the
freez i ng and thawing of mu tton carcasses the s i z e and shap e i n forma t i on
i s of ten l i mi ted to the relevant we i ght range ( Earl e & Fleming 1 9 67 ,
Van i chseni e t al 1 97 2 ) . This i n format ion i s generally not suff i c iently
de tai led to test pred i c t ion me thods o ther than product spec i f i c
emp i r i ca l formulae .

The restr i c ted na ture of publ i shed data l ed to the formula t i o n of the
f i rs t two a i ms of the pr esent research :
( 1 ) to collec t thaw i ng data f or the b as i c s l ab , infi n i t e cyl i nder ,
sphere and r ec tangular b r i ck shapes
( 2 ) to coll ec t fr eez i ng and thawi ng data f or other ( ir r egular) shap es .
Resea rch Obj ec t i v es 34

T h i s expans ion of the data b ase was necessary to fac il i tate tes t i ng of
thawi ng t i me p red i c t ion me thods for s i mp l e shapes and pred i c t i on
methods app l i cab l e to i rregular geometr i es for both freez i ng and
thaw i ng .

Tab l e 3 . 1 g ives the des i rab l e attr i bu tes of any method for food
freez i ng and thawi ng t i me pred i c t ion d i scussed by Slatter & J ones
( 1 97 2 ) , Cl eland ( 1 97 7 ) and Cl eland & Ear le ( 1 9 77 a) . The methods
ava i l able are conv en i ently d iv i ded i nto two broad groups for fur ther
study :
Grou p ( I ) : Formulae requ i r i ng on ly hand calcul a t i on
Group ( II ) : Numer ical me thods requ i r i ng computer calcul ation.
These groups d iffer s l i gh tly from those u sed in Sec . 2 . 3 and 2 . 4 . Many
e ng i neers do not have the sp ec i al ised knowledge to impl ement nume r i cal
method s . They will not u se Group I I me thod s unl ess they ar e ava ilab l e
as computer packages . Such people therefore mus t often rely on the
simple formulae ( Group I) . The g rea ter soph ist i cation of numer i cal
me thods may jus t i fy the ir u se if comp l ex cond i t ions are to s tud ied and
more det a i l ed des i g n i n forma t i on is requ ired . Often imprec i s e
k nowledge o f the pha se change cond i t i ons and uncerta i nty in thermal
data me an that the more comp l ex numer i ca l methods are no more accurate
in p rac t ice than Gr oup I me thod s .

Cond i t ions usually imposed i n the der i v a t i on of Group I methods are


l i s ted in Tab l e 3.2. If any of these cond i t ions ar e v iolated Group I
methods can on ly b e app l i ed to p rac t i ca l p robl ems by tak i ng app rop r i a t e
av erages f o r the phase change p rocess . I n cont ras t Group II me thods
are app l i cab l e whe ther these assump tions are made or not .

Pr ov i ded the e ight assump t ions in Table 3 .2 are met , freez i ng or


thaw i ng t i mes are i nfl uenced by the sev en major fac tors l i s ted in Tab l e
3.3. The accuracy and u sefulness o f any experi men tal data col l e c t ed i s
dep endent on the ab i l i ty to measure these parameters , t o control
accurately the exper i mental env i ronment and to mee t the cond i t io ns in
Tab l e 3 . 2 . The data se ts collec ted were desi gned so that all imp or tant
parame ters c ould b e exam i ne d . These data enabled the th i r d , f our th a nd
f i f th a ims of the current study to b e c on s i dere d . These were to:
( 3 ) ass ess the accuracy of b o th Gr oup I and Gr oup II me thods for
Research Ob j e c t i ves 35

pred i c t ion of thawi ng ti mes for regular shapes


(4) a ssess the accuracy of Group II me thods for p rediction of freez i ng
and thawi ng ti mes for i r r egular shap es and
( 5 ) i nv est igate the p oss i b i l i ty of develop i ng a Group I method to
p red i c t freez i ng and thawi ng t i mes for i r r egular shapes .
A dec is ion was mad e to l im i t the current study to si tua t i ons where the
e ight cond i t ions l i s ted in Tab l e 3 . 2 app l y . Th i s dec i s ion was ma de
b ecause of time and resou rce l i m i tat ions . Desp i t e th is rest r i c t i on a
w i de range of p rac t i ca l ly i mportant problems can s t i l l b e c overed .
Fu ture work may inv est igate si tua t i ons no t l i mi ted by these
assump t ions .

Tab l e 3 . 1 Desi rab l e Attr ibu tes of Freez ing and Thaw ing T i me Pred i c tion
Method s

( a ) Suff i c ient accuracy ov er a wi de rang e of cond i t i ons


( b ) S i mp l e and che ap l y processed cal cu l a t i ons
( c ) A pp l i cab i l i ty to a w i de rang e of b iolog i cal mater ials of v a r i ou s
s i z es and shap es
( d ) A pp l i cab i l i ty to the p rac t i cal l y impor t an t th ird k ind of boundary
cond i t ion in s i tua t i ons where there is superhea t i ng or subcool i ng
( e ) Mi n imal need for deta i l ed thermal p roperty data
( f ) Pred i c t ion of hea t f l ow and temp era ture prof i l es as func t ions of
t ime
( g ) Use of a un i f ied app roach for b oth freez i ng and thaw i ng .
( h ) Appl i c ab i l i ty where b oundary condi t ions are t i me and pos i t ion
var i ab l e
Research Obj e c t i ves 36

T ab l e 3 . 2 Cond i t ions R equ i red For Der ivat ion of S i mpl e Fr eez ing and
Thawing T i me Pred i c t ion Formulae

( a ) That the boundary and i n i t ial cond i t ions are constant wi th t i me


and /or p os i t ion
( b ) Homogeneous mater ials
( c ) That the th ird b ou ndary cond i t i on desc r i b es the heat transfer at
the b oundary adequately and that rad iat ion ef fec ts ar e
i ns i g ni f i cant
( d ) No i n ternal heat generat ion
( e ) I sotrop i c mater i a ls
( f ) N eg l i g i b l e dens i t y chang e dur i ng phase chang e
( g ) I n ternal heat transfer by conduc t ion only
( h ) No mass transfer , such as evap ora tion , at the surface
( i ) That the obj ec t reta ins i ts phys ical i n t egri ty dur i ng the phase
chang e p rocess .

Tab l e 3 . 3 Fac tors Affec t i ng Fr eez ing and Thaw i ng T imes

( a ) Thermal p roper t i es of the ma ter ial


( b ) S iz e of the obj ec t
( c ) I n i t i a l temp erature of the mater i a l
( d ) The amb i ent hea t i ng or cool i ng temp erature
( e ) The surface resi s tance to heat transfer as defined by the surface
heat t r ansfer coeff i c i ent
( f ) The geome tr i c con f igura t i on o f the obj ec t
( g ) The f inal temperature at the thermodynami c centre at the comp l e t i o n
of the phase chang e p rocess .
37

4 NUMERI CAL MET HOD FORMULAT I ONS

I t was neccessary to develop some numeri cal free z i ng and thawi ng t ime
pred i ct i on methods beyond what has been reported in the l i t erature i n
order t o meet the f our th r es ear ch obj ect i ve defi ned i n Chap . 3 .

4.1 INTRODUCTION

Numer i cal methods to pre d i ct rates of phase change processes in f oods


can be d i v i ded i nto thr ee groups :
( a ) the f i n i te el ement method
( b ) the f i n i t e di ff erence method
( c ) the approx imate ana l y t i cal methods tha t requ i r e numer i ca l
i ntegration o r ar e too compl e x f o r hand calcul a t i on .
The l at ter group have f ew advantages over the other two and have the
maj or d i sadvantage that usually severely l i mi t i ng phy s i cal
approx imat i ons are made . Ther ef ore th i s group of numer i cal methods was
not cons i dered further . Only numeri cal me thods that account for phase
change by cons i de r i ng con t i nuousl y tempera tur e var i abl e thermal
proper t i es were cons i dered as they are phys i cally the most real i s ti c .
T he al terna t i ve numer i cal methods trea t i ng phase change as a mov i ng
boundary probl em by assum i ng a uni que phase change t emperature do not
approx i mate phase change in foods very c l osely and ar e ther efore
i nf e r i or ( Cl el and 1 9 77 ) .

Advances i n compu t i ng t echno l ogy have meant that cos ts f or computer


computat ion power and dat a s torage ar e now rel a t i vely low so numeri cal
methods become attrac t i ve compared wi th s i mpl e analy t i ca l methods . The
ava i l ab i l i ty of comprehensi ve programs is i ncreas i ng so detail ed
knowl edge of comput er programm i ng and of the numer i cal techn i ques are
not neccessar i l y required to use thes e methods .

N umer i cal methods have been advocated as a s tandard metho d for


pred i c t i o n of phase change aga i ns t whi ch al l other pre d i c t i o n me thods
shoul d be compared , because numer i ca l methods are the c l osest to an
" exact" pred i c t ion method that e x i sts ( Hel dman 1 98 3 ) . · C ompari sons w i th
numer i cal methods shoul d only compl ement , not repl ace , compar i sons wi th
N umeri cal Method Formul a t i ons 38

actual exper i mental freez i ng and thaw i ng data ( C l eland & E ar l e 1 98 4 a ) .


A numeri cal method can only be accurate if it has been correctly
formul at ed and , impl ement ed and i f rel i ab l e data are used . C ompa r i sons
wi th e xper iments ar e necessary to assess these factors . The f i ni te
d i f f erence method i s l i mi ted by prac t i ca l cons tra i nt s to use for
regul ar shaped obj e c ts such as slabs , i nf i ni t e cyl i nders , s p heres ,
i nfi ni t e rods and rec tang ul ar bri cks . For these shapes f i x ed regular
gri ds can be us ed for the f i ni te d i ff er ence appro x im a t i o ns and
comparat i ve l y s i mpl e schemes developed . For i r regul ar shapes mor e
compl e x f i ni te d i ff er ence schemes and i rregul arly spaced gri ds c an b e
used ( F l emi ng 1 97 1 c ) . These ar e more compl ex to impl ement o n computers
than f i ni t e el ement me thods .

Numer i cal methods requ i re no physi ca l assump ti ons about the phase
change process but mak e numeri cal approximati ons . For complex bo undar y
cond i t i ons , i rr egular shapes , non- homogeneous mater i als and where
t emperature prof i l es change rapidly wi th posi t i on , the f i ni t e el ement
method is abl e to make mor e exac t approx imat i ons than the f i n i t e
d i fference me thod . F o r these reasons only the f i ni t e el ement method
has been used extens i ve l y for i rregular shapes ( Comi ni et al 1 97 8 ) .

4.2 THE F INITE ELEMENT METHOD

Al though the f i ni t e e l ement method has been wi de l y used for phas e


change ( C om i ni et al 1 978 ) none of the publ i shed appl i cat i ons of the
f i ni t e el ement method to phase change cons i dered three-d imens i onal
probl ems , and only l i mi ted tes t i ng of some two- dimens i onal f i ni te
element codes aga i ns t exper imental data has been car ri ed out
( Sec . 2 . 4 . 4 ) . A s i mpl e probl em spec i f i c program was avail abl e f or
three- d imen s i onal hea t transfer ( L im 1 97 5 ) but a general f i n i t e e l ement
program was not . Therefore a general formul a t i o n for heat cond uc t i on
i n three-d imensi onal geometry was devel oped and programmed ( C l eland et
al 1 98 4 ) .
N umer i cal M et hod Formul at i ons 39

4.2.1 F i ni te E l ement Method Formul a t i on

The f i n i t e el ement me thod can be formul ated for heat tr ansfer wi th


change of phase in a thr e e- d i mens i onal obj ect by usi ng the Gal er k i n
we i g hted resi dual method for spacewise di scret i zat i on o f E qs . ( 2 . 1 ) and
( 2 . 8 ) ( Z i enk i ewi cz 1 97 1 , Seger l i nd 1 97 6 ) . I ncorporat i ng the boundar y
condi t i ons ( E qs . ( 2 . 3 ) to ( 2.6) and ( 2.9) to ( 2 . 1 1 ) ) us i ng Green ' s
theorem , ye i l ds the n s i mul taneous d i f f er ent i al equa t i ons to be solved
impl i c i tl y for the t emperatur es at the n nodal poi nts defi ned
throughou t the reg i on . Wri t ten i n matr i x form the sol ut i on b ecomes
( Z i en k i ewicz & Parekh 1 9 70 , C omi ni et al 1 9 74b ) :

1: I(t) + c I(t) = f (4. 1 )


wher e I[ � thermal conduc tance mat r i x
I vec tor of noda l temperatures as a f unc t i on of t i me
c = thermal capac i t ance matr i x
!: thermal for c i ng vec tor
The t yp i ca l matr i x e l ements are:

(4.2)

(4.3)

(4.4)

whe r e W • 2 nr for cyl i ndri cal co- ord i nates


• 4nr 2 f or s pher i ca l co- ordi nates
,. 1 o therw i s e
shape f unc ti ons for the i th or j th node i n the e l ement
or s ur f ace undergoi ng i nt egra t i on .
Equa t i on ( 4 . 1 ) can be approx i mated by f i n i t e d i ffer enc es or f i n i t e
el ements and sol ved for f uture ti mes g i ven the appropri at e i ni t i al
condi t i on . U s i ng f i ni t e el ements in the f our th ( ti me ) d imensi on
greatl y i ncr eases the s i ze of the probl em for comput er i mpl ementat i on
but does all ow non- l i near es tima t i on i n the t ime doma i n . U se o f f i ni te
d i f f er ences i n the t i me domai n i s more common . The three t ime l ev e l ,
l i near , f i ni te d i fference scheme propos e d b y L ees ( 1 966 ) i s wel l proven
N umer i cal Method Formul ati ons 40

for phase change probl ems ( C om i n i & Bona c i na 1 97 4 ) :

(4.5)

I t has the advantage over other f i ni t e d i ffer ence schemes that it is


central l y bal anced s o only cent ral t i me val ues o f the � and C mat r i ces
and f: vec tor ar e requ i re d . Therefore i terat i ve t i me stepp i ng
algor i thms are not needed to i ncor porate correct val ues of the
t emperature and t i me dependent thermal proper t i es , and the boundary
cond i ti ons . T he L ees ' scheme has been shown to be uncond i ti onall y
st abl e and converg ent i n the cont ext of f i ni t e el ement anal ys i s ( C om i n i
et al 1 9 7 4 b ) .

By u s i ng the temperature- dependent apparen t spec i f i c heat capac i ty and


thermal conduc t i vi ty ( s uch as F i gs . 2 . 2 to 2 . 3 ) thi s nume ri cal method
wi l l cl os e l y mod el the true physi cal process dur i ng phase change if
data are accurat e . To account for both chang es i n thermal proper ti es
over each e l ement vol ume wi th t i me and the pos s i bl e t i me dependence of
the boundary condi t i ons , the i nt egra t i ons shown in Eqs . ( 4 . 2 ) to ( 4 . 4 )
must be repeated f or each t i me step , and I, C and f re- eva l uated . In
Eqs . ( 4 . 2 ) t o (4.4)
the summat i ons are t aken f or each e l ement , V e • i n
the e l ement reg i on , and f or each e l ement sur f ac e , S 2 e or S 3 e • on whi ch
E q . ( 2 . 3 ) or ( 2 . 4 ) is spec i f i ed . The spa t i al arrangement i s chosen so
that wi thi n each e l ement or on each sur f ace the var i abl es k , C , Q, $,

h con • E,
T ac and T ar ar e no l onger dependent on pos i t i on . I n phy si ca l
t erms , th i s means that no e l ement covers a reg i on made u p o f mor e than
one materi al type , and that boundary s urfaces are chos en so that
boundary var i ab l es $. h con • T ac and T ar can be ass umed cons tant f or
E,
that s urface , but not neccessar i l y the same as for other boundar y
s ur f aces . Knowl edge of these var i abl es i s often subj ect t o s i gn i f i cant
uncert ai nty . Therefore from a prac t i cal v i ewpoi nt the above
l im i tat i ons shoul d no t be restr i ct i ve . The dependence on t emperatur e
of k , C , Q , hrad • and on t i me of h con • �. T ac and T ar at each of the
s urfaces , r ema i ns .

T he f ul l numer i cal i nt egrat i on of Eqs . ( 4 . 2 ) t o ( 4 . 4 ) requ i res a l arge


c omput at ion effort a t each time s t ep . Therefore a s i mpl e r formul a t ion
is also proposed i n whi ch Eqs . ( 4 . 2 ) to ( 4 . 4 ) b ecom e :
N umer i cal M ethod Formul at i ons 41

J
aN j +
aN i aN j
- - -- W dV
ay az az
N j W dS 3 (4.6)

(4.7)

IS Ni w dS 2
ze
Is N i w dS 3 (4.8)
3e
T h i s requ i res the f urther assumpti ons that wi th i n each of the el emen t s
o r across each of the el ement s urfac es f o r wh i ch the summat ions are
made , that the proper t i es k , C , Q and h rad are cons tant wi th res pe c t to
temperature , but that these values can change from one t i me s tep to the
ne xt . Even though this is not physi cally accurate , if a val ue of these
proper t i es that is repres ent a t i v e of the whol e el ement or el ement
surface can be de termi ned , and if the e l ements ar e not l arge this
approach may be suff i c i ently accurate for many pur poses . G i ven that
the t i me s t ep i s control led so that temper ature changes per t i me s t ep
are small , the maj or di sadvantage of thi s s i mpl er formul a t i on is that
the proper t i es ar e not re- eva l ua t ed at each i nt egrat i on poi nt . I n the
full formul a t i on evalua t i on at each i nt egra t i on point hel ps accoun t f or
property vari at i on over the e l ement reg i on so that os c i l l atory
behav iour , espe c i al l y near the phas e chang e t emperatur es wher e thermal
proper t i es change rap i dl y , i s l ess l i kely ( C l el and et al 1 98 4 ) . The
advantage of the simpl er formul a t i on is that the i nt egra t i ons shown i n
Eq . ( 4 . 6 ) to ( 4 . 8 ) need only to b e calcul ated for the f i rs t t i me s t ep .
T hereafter the constant val u es resul t i ng from the i nt egra tion c an b e
stored and m ul ti pl i ed e l ement b y e l ement ( and s ur f ace by s ur f ace ) by
the appropriate proper t i es , and the s ummat i on compl et ed . T h i s requi res
f ar l ess computat i on t i me and computer memor y than f or the f ul l
f ormul a t i on .

The cho i ce of formul at i on must take i nto account the t rade- off between
loss in acc uracy and sav i ngs in comput at ion cos ts . As c omput a t i on
costs are usually f ar l es s than the cost of data preparati on ( S egerl i nd
1 97 6 ) , the sav i ng s may be qui t e smal l . Also uncert ai nti es in thermal
propert i es and bounda r y cond i ti on dat a are s ubstant i al ( Sec . 2. 6 ) .
Provi ded numeri cal errors are small compared wi th physi cal
uncer t a i nt i es no wort hwh i l e i ncrease i n pred i ct i on accuracy w i l l occur
N umeri cal Method Formul a t i ons 42

b y mak i ng t he method more soph i s t i ca t ed . T he s i m pl i f i ed method uses


thi s fact and poss i bl y increases numeri cal error , but may not l e ad to
s i gn i f i cant l y poor er pred i ct i ons of exper imental va l ues . If there was
a si gni f i cant i ncreas e in error i t woul d be e xpected to b e a maximum
when the t emper atur e grad i en t s across the el ements were l argest i n the
t emperature range where phas e change occurs . In thawi ng , this
corres ponds to h i gh sur f ace heat transf er coef f i c i ents , i ni t i al
t emperatur es clos e to the phase change t emperatur es and amb i ent
temper at ur es great l y d i f f erent from the phase change temperat ures
( Cl el and et al 1 9 8 2 ) .

4.2.2 C ompu ter Impl em entat i on

Both the f ormul at i ons g i ven by Eqs . ( 4 . 1 ) to ( 4 . 8 ) were programmed in


F or tran 7 7 . The full cod es are g i ven i n App . D whi ch also i n cl udes
da ta pr eparat i on not es . The p rograms use numer i cal i nt egrati on of the
el ement mat r i c es and a Gauss i an el i m i na t i on al gori thm for symme t ri c ,
pos i t i ve def i n i t e , banded mat r i ces wi th back subst i tut i on as the
equa t i on sol ver . Most of the computer pro cess t i me to run these
programs i s due to es t i mat i on of thermal proper t i es and numer i cal
i nt egrat i on of the el ement mat ri ces . T herefore us e of a mor e
soph i st i cat ed and eff i c i ent techni que to solve the s yst em of e quati ons
i s not j us t i f i ed . C omputer memory requ i r ements ar e h i gh b ut can b e
mi nimi sed b y mat ch i ng array d imens i ons c l osely t o the probl em s i ze or
by usi ng peri pheral s torage dev i ces . The speci al featur es des cr i bed
below are i nc l uded i n the programs .

The Lees ' scheme makes a l i near appro ximat i on by f i ni te d i f f er ences in


the i nt erval ( t- � t ) to ( t + � t ) . For a l i near var i at i on i n t emperatur e
the rad i a t i on hea t transf er coeffi c i ent w as accurately approximat ed by
the method o f C om i ni et a l ( 1 974 b ) :
£ a ( ( T + T �r ) ( ( T ) + ( T � ) ) + ! ( T � -T � ) ( T � +T - �T
h rad •
i i 2 2 1 1 2 i i-1 )
+ � ( T �; 1 -T �� l ) ( T i -T i - l ) ( T i - l +T � l ) + ( T i -T i - l ) 2 ( T i + �T � 1 - � T � 1 ) ] (4.9)
i n which only k nown val ues of t emperature are used . T h i s all ows
central val ues of t to be used wi th good accuracy and an i terat i ve
procedure ( Fr i v i k et al 1 97 7 ) to solve Eq . ( 4 . 5 ) i s not r e qu i re d . As
r ad i a t i o n i s not a maj or form of heat transfer d ur i ng food phase change
processes thi s method shoul d be suff i c i en t l y accurate for mos t
N umer i cal M ethod Formul at i ons 43

pur poses .

I n f i ni te el ement analysi s stab l e osc i l l at i ons frequently occur


( Y al amanch i l l i & Chu 1 97 3 , Myers 1 978 ) . They can be accentuat e d when
us i ng central values of [ ( Comi ni & Lewi s 1 97 6 ) . U pdat i ng t emperatures
at the end of each t i me step usi ng the averag i ng procedure (Wood &

L ew i s 1 975 ) :
Ti-1 � ( T i - 1 +T i +T i + 1
= ) (4.10)
or reduc i ng the t i me step ( Segerl i nd 1 9 76 ) wi ll reduce these
os c i l lati ons . Too small a t i me s tep is undes i rabl e as computat i on
t i mes are i ncreased and computer round i ng errors can be come
s i gn i f i cant .

The accuracy of Eq . ( 4 . 1 ) wi ll be best i f the vari at i on of k, C, Q,

h rad • h con • �� · T ac and T ar wi th t i me over each t i me step is small or i s


reasonably l i near . For mos t probl ems h con • <f> , T ac and T ar ar e
ess en t i al l y cons tant wi th ti me , so that t emperature dependenc e of k , C ,
Q and h rad wi l l be the l arger contri but i on to i naccurac y .

A number of techni ques have been used t o f i nd values of k , C and Q that


are represent a t i v e when rates of change of these quanti t i es wi th space
ar e l arge , such as near the phase change temperatur es ( Sec . 2 . 4 . 4 ) .
T hes e approaches may l ead to erroneous r es ul ts (Morgan et al 1 97 8 ) so
d i rect eval ua t i on of these pr oper t i es as f uncti ons of temperat ur es i s
us ed . Control of the t ime s tep , so that excess i ve t emper ature changes
over each t i me s t ep cannot occur , i s the ul t i mate strategy t o g i ve
acc uracy . The t i me s tep control ens ur es proper heat bal anc es a cross
el ements and boundar i es , helps reduce numer i cal os c i l l at i on and
therefor e i n creas es prec i s i on in the solution. Smal l t emper ature
changes per t i me step r educe the error i n ass umi ng that the r e i s l i near
var i at i on i n k, c, Q and hrad i n the i nt er val ( t- 6 t ) to ( t+ 6 t ) .
T herefore the use of val ues f or K, C
and [ at the central t i me val ue ,
t, wi l l be an accur at e approximat i on . An automat i c t i me s t ep
adj ustment ( C om i n i et al 1 97 4b ) i s us eful to keep tempera t ur e changes
per t i me step l ow wi thout great l y i ncreas i ng comput at i on t i me . This i s
ach i eved b y choi ce o f val ues f or the parameters 6Tmi n an d
appropri at e
6Tmax • whi ch are the maximum and m i n i mum nodal t em per ature changes per
t i me step and ar e used to dec i de when to adj us t the t ime s tep . The
N umeri cal M ethod F ormul a t i ons 44

mos t appr op r i at e si zes of �T mi n and � T max are gover ned by the wi dt h of


the peaks in the p rof i l es of k , C and Q v ersus temperature (Morgan e t
a l 1 97 8 ) .

The accuracy of thermal proper t y approximat i ons and the numer i ca l


method appro x ima t i ons c an b e chec ked b y cal cul a t i ng the heat bal anc e :

apparent i nt ernal
entha l py chang e
heat fl ow i nt ernal
in through + heat
the s ur f ac e genera t i on
dV] dt

= 1 . 0 for bal ance (4.11 )


A heat balance does not occur if proper t y approximat i ons ar e
i nsuf f i c i ently accurate , time step or s pa t i a l i nt ervals ar e too l arge
or the numer i cal approx imat i ons used are not val i d . For exampl e , a
heat bal ance greater than 1 . 0 generally i nd i cates that " j ump i ng" of the
l atent heat peak has occurred ( C l el and & Ear l e 1 97 7 b ) and that t i me
s t eps shoul d be shor tened by reduc i ng both �T mi n and �T max · S imi l ar l y ,
a heat balance l ess than 1 . 0 general l y indi cates that t emperatur es at
for some el ements have " stayed on" the l at ent heat peak too l ong , and
that t i me s teps and/ or s pace i ntervals shoul d be reduced . T he f i rst
case of " j umpi ng " of the l atent heat peak i n the apparent s pec f i c heat
capac i t y prof i l e is a probl em in free z i ng calc ul at i ons because as the
temperat ur es drop b e l ow T i f the peak i s suddenly encounter ed , wher e as
the s econd case occurs mos t freque ntl y i n thawi ng cal cul at i ons because
the l atent heat peak suddenly d i sappears as t emperatur es ri se above
Tif .

T he programs can be used f or heat transf er probl ems i nvo l v i ng obj ects
of one , two or three- d imensions . T he a x i s ymmetri c featur e can be used
to s i mpl i f y probl ems wher e one or more axes of rota t i onal s ymmetry
exist . The el ements used ar e based on the rect angul ar "S erend i p i ty"
fami l y ( Z i en k i ewi cz 1 97 1 ) . L i near or quadrat i c i s oparametr i c and
super parame t r i c el ements are ava i l abl e . The high ord er el ements r educe
the total number of e l ements necessary and a l l ow curved e l ement
boundari es to b e use d t o model i rregul ar geomet r i es . The only s pe c i f i c
use of s ubparametr i c e l ements woul d be f or regul ar shapes wher e a
quadr at i c t emperature prof i l e i s required . A s quadra t i c i so parametri c
el ements ar e capabl e of handl i ng thi s case j us t as wel l , wi thout
Numer i cal Method F ormul at i ons 45

s i g ni f i cant i ncreas es i n computat ion t i mes , subparametri c el ements are


not i nc l uded i n the programs .

A l l the boundary cond i t i ons g i ve n i n the f ormul at i on ( Sec . 4 . 2 . 1 ) can


be used , and any i ni t i al t emper ature d i s t r i b u t ion may be s pec i f i e d .

4.2.3 F i ni te E l ement Method Tes t i ng

Tes t i ng of the f i ni t e e l emen t codes was made aga i ns t exper i mental


f ree z i ng dat a and anal y t i cal sol ut ions f or s impl i f i ed phase change
probl ems ( Cl e l and et al 1 98 4 ) . The number of no des and e l emen t s used
was a compromi se between the hi gher accuracy and l ess frequent
occurrence or smal l er ampl i t ude of os c i l l at i ons ach i eved wi th a l arger
number of nod es , and pra cti cal l i mi t at ions on the comput at ion t ime and
computer memor y storage requ i rements ( Seger l i nd 1 97 6 ) . S imi l arl y , the
number of i nt egration poi nts us ed in the numeri cal i ntegr ation of
el emen t mat r i ces was a balance between accuracy and compu tat i on time.
The us e o f thr ee i nt egration poi nts i n each d i rection was found t o g i ve
adequa t e accuracy in mos t cases . For the f ul l formul at i on the
c om b i na ti on of number of nodes and type of el ements ( l i ne ar or
quadrat i c ) was found to be l ess important as l ong as abou t e l ev e n nodes
( and consequen tly ei ther ten l i near el ements or f ive quadr ati c
el ements ) were used i n each d i rect i on ( C l el and et al 1 98 4 ) . For the
s impl er formul ation the number of el ements i s mor e cri t i cal as the
assumpt i on of cons tan t k , C and Q over each el ement becomes l ess
accurate for a f ew l arge quadrati c el ements rather than wi th more ,
small er l i near el ements . The benef i t of be i ng abl e to use a l ower
n umber of quadrat i c el ements is ther efore l essened because the t i me
step mus t be decreased to mai nta i n heat balances . I ncreas i ng the
n umber of nodes or el ements f or the full formul a t ion or the number of
i nt egra t i on poi nts f or both formul ati ons , above these s ugges t ed values ,
gave only mi nor changes i n pred i ct i ons . Quadrat i c el ements shoul d be
used i f the p roduc t s ur f ace i s curved or i rregul ar .

The exact sol ut i on due to Neumann ( Carsl aw & J aeger 1 95 9 , p . 2 8 2 ) f or a


spec i f i ed temperatur e boundary cond i tion , the analy t i cal sol uti on f or a
spec i f i ed s urface heat flux ( Carsl aw & J aeger 1 959 , p .75) and a
numer i ca l sol uti on f or the rad i at i on boundary cond i t i on ( Haj i -Shei kh &
N umeri cal Method F ormula t i ons 46

Sparrow 1 96 7 ) were used to tes t the programs f or the range of boundary


condi t i ons i n corporated . S im i l ar accuracy to that shown for other
f i ni te el ement codes ( C om i ni et al 1 97 4 b ) was demons trated ( Tabl es 4 . 1
to 4 . 3) . Both f i ni t e el ement formul a t i ons gave c omparabl e
temperature/ t i me prof i l es wi th those g i ven by the f i ni te d i f ference
method for the same probl ems . F or heat conduction in sol i ds wi th the
th i rd k i nd of boundar y cond i t i on the onl y solut i on for
thre e- dimens ional geometry i s due to Newman ( 1 936) . F i ni t e el ement
pred i ct i ons usi ng a coarse 6x6x6 node gri d ( Tabl e 4 . 4 ) gave simi l ar
pred i ct i on accuracy to a f i ni t e d i ff erence program for the i denti cal
probl em ( C l el and & Ear l e 1 97 9b ) . The thermal properti es are cons tant
in th i s l as t probl em and i t is not d i f f i c ul t to sol ve the probl em
numer i cal l y . Ther efore th i s probl em i s mor e a check that the method
has been impl ement ed correctly on the com put er , rather than a check
that the method is accurate in general .

C omprehens i v e experi ment al data av a i l abl e for heat cond uc tion wi th


phase change were l i mi ted to freezi ng of the regul ar shapes namel y :
sl abs ( Cl el and & Earle 1 97 7 a ) , i nf i ni t e cyl i nders and spher es ( Cl el and
& Ear l e 1 97 9 a ) and rect angul ar bri ck s ( C l el and & Ear l e 1 97 9 b ) , s ubj ec t
to the boundary condi tion of the third k i nd . Al though the full
f ormul at i on coul d not be run on all the rec tangul ar bri ck data due to
e xcessi v e comput at i on t imes , the overall predi ct ion accuracy compared
favourabl y wi th the f i ni te d i f fer ence pred i ct i on of this data ( C l el and
et al 1 98 4 ) .

I n al l test i ng the s impl i f i ed formul at i on gave s i mi l ar resul ts to the


f ul l f o rmul a t i on though often small er t i me steps wer e requ i red to
achi ev e heat bal anc es f or quadrat i c el ements . C omputat i on t i mes f or
the three- dimens ional probl ems wi th 2 1 6 nodes i n a 6x6x6 g r i d of 1 25
evenly s i zed , l i near i s opar ametr i c el ements over 250 t i me s teps were
about 1 5 000 sec . and 3 0 00 sec . f or the two formul ations r es pe c t i v el y
o n a Prime 750 compu t er . By compar i son , the thr ee- d i mens i onal , L ees '
s cheme , f i ni te d i fferen c e method program wi th a 6x6x6 node g r i d used
about 300 sec . computer process t i me f or 250 time steps on a Prime 750
c omputer .
Numer i ca l Me thod Formulat i ons 47

Tab l e 4 . 1 Compar i son of Resul ts From the F i n i te El ement Me thod Programs


W i th Neumann ' s Solut ion 1 For Thawi ng of a Slab Sub j e c t to the
F ir s t K i nd of Boundary Cond it ion

D 1 . 0 m , ks kl = = 1 . 0 W m - 1 °C - 1 , C S c
1
� 1 . o MJ m -3 oc - 1 ,
L 3
75 . 0 MJ m- , T a = 1 5 °C , T in - 1 5 °C .
=

Interface t neum 2 t femf 2 t fems 2


Pos i t ion
(m) ( s ec ) ( s ec ) ( s ec )

0 . 05 1 0 400 1 0 800 9700


0.10 4 2500 42700 3 8600
0.15 93 400 95900 88900
0 . 20 1 6 6000 1 70 800 1 6 2800
0 . 25 25 9300 26 4700 258000

1 Car slaw & Jaeger ( 1 959 . p . 282 )


2 t neum t i me calcul ated from Neumann ' s so l u t i o n 1 , t femf
= t ime =

calculated from the ful l f i n i te el ement formulat i on ( Eqs . ( 4 . 2 ) to


( 4 . 5 ) ) and t fems t i me calculated from the s i mp l er f i n i t e e lement
=

f ormul at ion ( Eqs . ( 4 . 5 ) to ( 4 . 8 ) ) . Latent heat was evenly


d i str i buted ov er the temper ature range from -0 . 25 °C to 0 . 25 °C , 21
nod e s and 20 l i near i s o pa r a me t r i c e l em e n t s w er e u s e d in th e f i n i te

e l ement programs .
Numer i cal Method Formulat ions 48

Tabl e 4 . 2 Comparison of Resul t s From the F i n i t e El emen t Method Programs


W i th a Known Ana l y t i cal So lut ion 1 For Cool ing of a Slab
Subj ect to the Second K i nd of Boundary Condi t ion

0 = 1 . 0 m, k = 1 . 0 W m - 1 oc - 1 , C = 1 . 0 MJ m -3 °C - 1 , 4> = 1 0 0 W m -2 ,
T i n = 1 00 °C . Values of tempera ture ( °C ) are tabu lated.

Slab Sur face At 0 . 1 m From Sl ab Sur face

T ime Ana l y t i cal Numer i ca l 2 Analy t i ca l Nume r i cal 2


( sec )

20 84 .04 83 . 83 92 . 09 9 2 . 90
40 77 . 4 3 77 . 28 86 . 0 4 85 . 90
60 72.36 72.24 81 . 22 81 10

80 68 . 09 67 . 98 77 . 09 76 . 99
1 00 64 . 3 2 64 . 22 73 . 43 73 . 3 4

1 Carslaw & Jaeger ( 1 9 59 , p . 7 5 )


2 Both the ful l formul at ion ( Eqs . ( 4 . 2 ) to ( 4 . 5 ) ) and the s i mpl i f ied
formulat ion ( Eqs . ( 4 . 5 ) to ( 4 . 8 ) ) gave i dent i cal r esult s , 2 1 nodes
and 20 l i near i soparametr i c elemen t s were used in the f i n i te e l emen t
programs .
Numer i ca l Method Formulat ions 49

T ab l e 4 . 3 C ompar i son o f Results From the F i n i te Elemen t Method P rograms


W i th a Numer i c al Sol u t ion 1 For Coo l i ng of a Slab Sub j e c t to
the Th i r d K i nd of Boundary Cond i t ion Includ ing Rad i a t ion

1 1
D 1 .0 m, k 1 . 0 W m- °C - , C "' 1 . 0 MJ m - 3 °C - 1 , h = 1 .0 W m -2 oc - 1 ,
£ = 1 . 0 , Tac 0 K , T ar = 0 K. Values o f Y ar e tabul ated .

Slab Sur face Slab C entre

Fo Numer i c al 1 F EM 2 Nume r i cal 1 F EM 2

3
£ oT i n D / k

0 . 02 0 . 80 0 .79 1 . 00 1 . 00
0 . 06 0 . 70 0 . 70 1 . 00 1 . 00
0 . 20 0 . 58 0 . 58 0 . 94 0 . 93
0 . 40 0 . 50 0 . 50 0 . 80 0 . 80
1 . 00 0 . 33 0 . 32 0 . 50 0 . 50

2 . 00 0.16 0.15 0 . 23 0 . 23

3
£ oT i n D / k 4

0 . 02 0 . 70 0 . 69 1 . 00 1 . 00
0 . 06 0 .61 0 .61 1 . 00 1 .00
0 . 20 0 .5 1 0 .5 1 0.91 0.91
0 . 40 0 . 43 0 . 43 0 . 76 0 .7 6
1 . 00 0 . 29 0 . 28 0 . 44 0 . 45
2 . 00 0. 1 3 0 .1 4 0 . 20 0 .21

1 Haj i -She i kh & Spar row ( 1 9 6 7 )


2 FEM "' calculated from the e l ement method , both the full formul a t ion

( Eq s . ( 4 . 2 ) to ( 4 . 5 ) ) and the s i mpl i f i e d f ormula t ion ( Eqs . ( 4 . 5 ) to


( 4 . 8 ) ) gave i dent i ca l resul ts .
Nume r i cal Method Formulat ions 50

Tab l e 4 . 4 C ompa r i son of Results From the F i n i t e Element Me thod P rograms

W i th a Known Analyt i cal Solut ion 1 For Cool i ng of a Cube


Sub j e c t to the Th i r d K i nd of Boundary Cond i t ion

Bi 4.0. Values o f Y are t abulated .

C entre of Face Centre of C ube

Fo Ana ly t i cal 1 Nume r i cal 2 Ana ly t i ca l 1 Numer i ca l 2

0 . 04 0 . 489 0 . 495 0 . 999 1 .001


0 . 08 0 . 3 87 0 . 3 92 0 . 974 0 . 985
0.12 0.318 0.31 9 0 . 897 0 . 91 2
0.1 6 0 . 26 1 0 . 265 0 . 7 90 0 . 80 4
0 . 20 0.215 0.218 0 . 6 77 0 . 6 88
0 .24 0 . 1 77 0 . 1 80 0 . 570 0 . 580

1 N ewman ( 1 9 3 6 )
2 Both the f u l l formu lat ion ( Eqs . ( 4 . 2 ) to ( 4 . 5 ) ) and the s i mp l i f i ed

formul at ion ( Eq s . ( 4 . 5 ) to ( 4 . 8 ) ) gave i dent i cal r esul t s , a 2 1 6


( 6x 6 x 6 ) node gr i d and 1 2 5 l i near i soparame t r i c e lements were used i n
the f i n i t e e l ement pr ograms .
N umer i cal Method F ormul at i ons 51

The ab i l i t y of the code t o deal wi th c urved i rregul ar g e ometry i n two


d i mens i ons was conf i rmed by modell i ng the cyl i ndr i ca l shape w i th the
g r i d g i v en in Fig . 5 . 1 5 . P redi c t i on accuracy was s i m i l ar to that
obta i ned usi ng r egu l ar one- d i mens i onal ax i s ymmet r i c el ements . A l though
the f i ni t e el ement code h as not been proven for a truely i rr eg ul ar

shape , there is no reason why simi l ar p r ed i ct i on accuracy ca nnot be


a c h i e v e d i f s u f f i c i en t l y det a i l ed nodal gri ds are us ed , as the f i n i te
el emen t method does not treat regul ar shaped e l ements any d i f f er entl y
f r om i rregul arly shaped ones .

T he two f i ni te el emen t programs were ther efore cons i de r ed to be


acc urately formul ated and correctly i m pl ement ed . H enc e they w er e
ava i l ab l e as pred i c t i on methods tha t wou l d be expected t o g i ve a c c urate
pred i c t i ons of phase change in obj ects of i r r eg ul ar geome try .

D i s ag r e ement wi th e xper imen tal data is more l i k e ly to be due to


imprec i s e thermal and e xp e r i mental data than to i nade qua c i es in the
f o rm ul a t i on or prog ram . I f pred i c t i o n me thod error ari s es i t i s more
l i k e l y to be because prac t i cal l i mi ts on memory and compu ter pro ces s
t i me ha ve meant that coarse s pace and t i me g ri ds ha ve been u s ed .

4.3 THE F I N I T E DI FFE RENCE ME T HOD

Dev el opm ent of the f i n i te e l ement method has meant that the f i ni te
d i f f er ence me thod is mai nly us ed f or pre di c t i on of heat transf er i n
regul ar l y s haped obj ects . For probl ems w i th regul ar shapes , f in i t e
d i ff er en c es requ i r e l es s det a i l ed pr ogr am preparat ion and hav e l ower
compu ta t i on t i m es . Bonac i na & C om i ni ( 1 97 1 ) , C l el and ( 1 97 7 ) and
C l el and & E arle ( 1 97 7 a ) s t u d i e d f i ni t e d i f f er e nce methods and f ound the
thr ee t i me l evel i m pl i c i t L ee ' s s cheme t o be the bes t f or mod el l i ng
f r e e z i ng of food . C omp uter program s us i ng the L ee ' s scheme for
i nf i ni t e s l ab s ( C l el an d & Earl e 1 97 7 b ) , i nf i ni t e cyl i nders and s pheres
( C l el and & E arle 1 97 9 a ) , rect angular r o ds ( Cl el and 1 9 77 ) and
rec tangul ar b r i c k ( Cl el and & E ar l e 1 97 9b ) shapes are ava i l abl e and were
subsequently us e d . T hese programs have been shown to b e acc ur ate to
wi th i n about ± 1 0% compared w i th experi mental sl ab , i nf i ni t e c y l i nde r ,
sphere and r e ct angular b r i c k free z i ng data ( Cl el and et al 1 9 8 2 ) . T hey
also meet anal y t i ca l chec k s . T her e i s no r e as on why s i mi l ar pred i ct i on
a c c uracy c annot be expe cted us i ng these p rograms f or thawi ng of f ood .
N umer i cal Method F ormul at i ons 52

T he e qu i va l ent s chem e f or the f i ni te cyl i nder geomet r y has not been


publ i shed and is ther efore d es c r i bed her e .

4. 3 . 1 F i ni t e C yl i nder F i ni t e D i f f erence Scheme F ormul at i on

For the f i n i t e cyl i nder geometry the general heat co nduct i on equa t i on
is:

()T k ()T

[ :: ] [ :: ]
a a
c k + - - + k ( 4. 1 2)
at ar r ar ay
A l though the f i ni te cyl i nder i s thr ee- d i mensi onal , because i t h as an
ax i s of rotat i onal symm etry , i t can be cons i dered as two- d imens i o nal .
Bonac i na & C om i ni ( 1 9 7 3 ) hav e shown how the L ee ' s sc heme can be appl i ed
to probl ems in two d i mens i ons by usi ng s tandard a l ternat i ng d i r ec t i on
impl i ci t proc edur es . Equa t i on ( 4 . 1 2 ) can be appro x i mated by a L ees '
schem e :
i+ 1 i- 1
T j -T mj
ci . m
mJ
2H

+

k j
2m
m 1
+ +
( T i + 1 j -T i 1 j + T i + j - T i j + T i -
m- 1 m 1 m- 1
1 i- 1
m + 1 j -Tm- 1 j
)]
+
1 �
i i+ 1 i+1 i i i- 1 i-1
2 k mj + � ( T mj + 1 - T mj + T m j + 1 - T m j + T m j + 1 - T m j )
L
� 1 - T�j � 1 +T�j -T�j - 1 +T �j 1 -T�j � 1 )]
3 ( 6y )


+ k j � (T
-
( 4 . 1 3)

T a k i ng the radi al d i r ec t i on f i rs t , the two sweeps of the al t ernat i ng


N umer i cal M ethod F ormul at i ons 53

[
Second sweep ( i n y d i re c t i on ) :

i+1
- T mj -1
[
y ki
mj - i Jl + T mj r.l
i + 1 c i +Y ( i
mj
i
i mj - i
k mj + + k ) ] - T i+1
mj + 1
y ki
mj + l
J
i
= C mj i+1 *
T mj - y
�l mj
i i- 1
+ i ( T mj + 1 -T
i-1 )
mj -k
i ( i-1 i-1
mj - 1 T mj - T mj - 1
)] (4.15)

26t 26t *
where X = 2' Y = 2 and T are i nterme d i at e t emper at ures
3 ( 6r ) 3 ( 6y )
cal cul ated by the al ter nat i ng d i re c t i o n scheme that hav e no phy s i ca l
s i g n i f i canc e .

Because of the s ymmetry only a quadrant of the axi symmetr i ca l


two- d i mens i onal g r i d needs be cons i dered , wi th a s pace g r i d o f ( M + 1 ) x
( J + 1 ) nodes , whe re M6rxR and J 6y =D y / 2 . T he t h i rd k i nd of boundary
cond i t i on i s taken i nt o account in a s i m i l ar manner to that shown by
C l el and & Earle ( 1 97 7b ) for slabs and C l el and & Earl e ( 1 979a) f or
rad i al geometry .
A t the rad i al ce n t r e ( m = O ) the method of A l bas i ny ( 1 960 ) :
2
l i m 1 aT a T
(4. 1 6 )
a r .... a?
r +O ;

was used to av o i d the si ngul ar i t y . T he f i rs t sweep ( Eq . ( 4 . 1 4 ) )


becomes :

(4.17)

� �
set t i ng k -! j =k + ! j and T m- 1 j -T m + 1 j at al l t i me l ev el s . T he s econd
sweep i s unal tered f r om Eq . ( 4 . 1 5 ) .
A t the rad i al sur f ac e ( m=M ) :
aT
h ( T a -T w ) • k - (4.18)
or
N umer i cal M ethod F ormul at i ons 54

and the f i r s t sweep ( E q . ( 4 . 1 4 ) ) bec omes :

r�:1 ; [- x k �- lJ ] + r�1 * [c!u • x (k!-u•Mr)J


i i-i +
= C mj T mj
[ i
X k m- i i i-1 i-1
� j ( T m- 1 j -T mj + T m- 1 j - T mj )
] + i - i-1 )
X hAr ( 3T a- T mj T mj

i[ i i i- 1 i-1 i i i i- 1 i- 1
+ Y k mj + � ( T mj + 1 -T mj + 2T mj + 1 - 2T mj ) - k mj - � ( T mj - T mj - 1 + 2T mj - 2T mj - 1
)] (4. 19)

wh i l e the second sweep i s unal tered f r om E q . ( 4 . 1 5 ) . For both sweeps


the f ol l owi ng co r r ec t i ons to the thermal conduc t i v i ti es al ong the
s urface must b e made , due to the reduced area for conduc t i on :
.
1
M -tt C l ·
mJ "
2M
--

A t the surface i n the y d i r e c t i on ( j = J ) :


ar
h ( T a-T w ) = k - ( 4 . 20 )
ay
T he f i r s t sweep i s Eq . ( 4 . 1 4 ) and the second sweep i s una l tered f rom
for mak 1 ng C mj = C / 2 , k m+ � j = k m � j / 2 , k m- � j = k m- � j I 2 ,
. i i i i i i
Eq . ( 4 . 1 4 ) exc e p t mj +
because onl y hal f of the normal volume i s asso c i a ted wi th each s urface
i i+1 i i-1
node , and set t i ng k mj + � = hAy and T mj + 1 = T mj + 1 =T mj + 1 =T a .
A t cor ner s on the gr i d a comb i na t i on of both of the appropr i at e changes
t o the scheme g i ven abov e are used .

4.3.2 C omputer I mp l ementat i on and T es t i ng

T he f i n i t e c yl i nder f i n i te d i f f er ence p r ogram was checked aga i ns t the


k nown anal y t i cal sol u t i on for the th i r d k i nd of boundary cond i t i on and
constant thermal proper t i es ( N ewman 1 9 3 6 ) . T abl e 4 . 5 shows that the
agreement of t emperatur es for a 1 0 x 1 0 g r i d i s good . A l though no data
f or phas e change i n f i ni t e cy l i nders were av a i l abl e , f r e ez i ng data f or
i nf i n i t e sl abs ( C l e l and & E arle 1 9 7 7 a ) and i nf i ni t e c yl i nders ( Cl el and
& E ar l e 1 97 9 a ) were used to tes t the p rogram . Mak i ng e i ther the
cyl i nder l eng th d i mens i on ( for i nf i ni te cyl i nders ) or the radi al
d i mens i on ( f or i nf i ni t e sl ab s ) ver y l arge in compa r i s on wi th the other
d imens i on , heat trans f e r in these ext ended d i r e c t i ons becomes
i ns i gn i f i cant at the centre of the g r i d and the s i mpl er geomet r y was
c l os e l y approx imated . T he e xper i mental i nf i n i te s l ab f ree z i ng data was
pred i ct ed wi th a mean error of 0 . 8 % and 95 % conf i de nce bounds of ±1 2 . 2%
and the i nf i n i te cyl i nder exper iment a l f r ee z i ng dat a wi th a me an of
N umer i ca l M ethod F ormul at i ons 55

- 3 . 6 % and 9 5 % bounds of ± 9 . 6 % us i ng the thermal data g i ven in T ab l e


5.1 . I nd i v i dual pre d i c t i ons were i denti cal to the r es ul ts f or the
s pec i al i sed sl ab and i nf i n i t e cyl i nd er f i n i te d i f f er ence prog rams u s i ng
the same therm al data . I t was concl uded that th i s f i ni te d i f f e rence
scheme f or a f i n i t e c y l i nder was as accurate as equ i va l ent schemes f or
other r egular shapes .

I n the above s cheme i t was chosen to do the f i rs t sweep i n the rad i al


d i r e c t i on . The al t erna t i v e wi th the f i r s t sweep i n the y d i re c tion was
also de r i ve d . I t has the same trunca t i on errors as the s cheme wi th the

ra d i a l sweep f i rst and was found to no t g i v e si gni f i c antl y d i ff erent


resul ts ov e r the range of probl ems and d i ffer ent cyl i nd e r rad i us to
l e ng th rat i os used . C onsequen tly the f i rst scheme was ret a i n ed because
the rad i al d i mens i on is the more i m por tant in mos t p robl ems , al though
both are equal l y approp r i ate .

The p r og r am i nc l udes the heat bal ance an d tem per atur e updat i ng
techn i ques d i scuss ed i n Sec . 4 . 2 . 2 for f i ni t e el ement me thods , i n or der
to p r event osc i l l at i on and to detect l atent heat peak " j ump i ng "
p ro bl ems . T o prevent osc i l l a t i on d u r i ng calcul a t i o n of phase change

for i nf i ni te c y l i nders C l el and & Ear l e ( 1 97 9 a ) d e f ined :


i i i
k mj - 2l ( k m + ij + k m- ij )
-
(4. 21 )
The same metho d was used i n the f i n i t e cyl i nder p rogram .

F or eleven nodes i n each d i r e c t i on and 1 000 t i me s t eps a t y p i cal


computat i on t i me f or th i s scheme wri t t en i n F or t r an 77 was 2 5 0 sec . on
a P r i me 750 computer . By compari son , the f ul l f i ni t e el ement
f ormul at i on ( Eq s . ( 4 . 2 ) to (4.5) ) , u s i ng the same g r i d s i ze , g a ve
c omputa t i o n t i mes of about 1 0 00 s ec . for 1 0 0 0 t i me steps .
Numer i c a l Method Formulat ions 56 .

Tab l e 4 . 5 C ompa r i son o f Resul ts From the F i n i te C y l i nder F i n i te

D i f ference Me thod Program W i th a Known Ana ly t i cal So lut ion 1


For Cool i ng of a F i n i te C y l i nder Sub j ect to the Th i r d K i n d of
Boundary Cond i t ion

Or =0 . 02 m , Dy =0 . 02 m , k = 0.617 W m- 1 °C - 1 , C = 3 . 6 9 MJ m
- 3 oc - 1 •

An 1 1 x 1 1 node g r i d was used . Values of Y are tab ul ated .

Edge of F i n i te C y l i nder Cent r e of F i n i t e C y l i nder

T i me Anal y t i ca l 1 Numer i c al Ana ly t i cal 1 Numer i ca l


( s ec )

1 00 0 . 6 25 0 . 624 0 . 936 0 . 93 4
300 0 . 398 0 . 39 7 0 .626 0 . 6 25
500 0 . 260 0 . 259 0 . 40 9 0 . 407
700 0 . 1 69 0 . 1 69 0 . 266 0 . 265
900 0. 1 10 0. 1 10 0. 174 0 . 173

-1
h 1 000 . 0 W m - 2 oc

15 0 . 0 38 0 . 037 0 . 959 1 . 000


45 0 . 01 2 0.01 4 0 . 90 4 0 . 934
75 0 . 00 7 0 . 006 0 .71 8 o . 7 41
1 05 0 .004 0 . 00 4 0 . 524 0 . 54 1
1 35 0 . 003 0 . 00 3 0 . 370 0 . 3 83

1 Newman ( 1 9 3 6 )
57

5 EXP ERIMEN T A L PRO C EDUR E AND DA T A COL LEC T ION

5. 1 EXP ER IMEN TA L ERROR

Whenever any fr eez i ng or thaw i ng t i me pred i c t ion method is compared

wi th experi ment al freez i ng and thaw i ng dat a , s ome lack of agr e ement i s
i ne v i tab l e . Th i s imp rec i s ion may ar i se f r om one o f three sources :
( a ) unc erta i n t y i n th ermal p roper t y dat a for th e mat e r i a l be i ng fr oz en
or thawed ( data error )

( b ) imprec i s e knowl edge and co ntro l of the free z i ng or thaw i ng

cond i t i ons ( e xpe r i mental err or ) and


( c ) i naccuracy ar i s i ng from as sump t i ons or ap prox imat ions made in the
d er i va t io n of the p r ed i ct ion me thod ( p red i c t io n method error ) .

Thermal p roper t y data uncerta inty depends on th e ma te r i a l used for the


exper imen ts and is d i scussed in Sec . 5 .2. T o assess in i solat i on as
far as i s pos s i b l e the magn i t ude o f the th i r d sou rce of i naccuracy , the
a i m of any expe r i ment a l proc edur e or techn i q ue i s to k eep e xpe r i me nta l
errors r andomly d i st r i buted and sma l l i n s i z e .

The f i r s t o f two s our ces o f experi men tal error is th e err or a r i s i ng


from imperfec t control o f ex per imental cond i t i o ns . Many var i ab l es are
contro l l e d to pre-se t va l ues . There i s cont ro l error w i th i n e ach run
as well as control p roblems i n atta i n i ng the same pre - set v a lue in a l l
s i mi lar runs . Th e ab i l i ty t o reduce the con trol err o r c an b e me asured
from the var i ab i l i ty of rep l i cate runs for the same set of nominal
e xper imental cond i t io ns .

Second ly , there i s the error ar i s i ng from imp rec i s e knowl edge of the
expe r i mental cond i t ions . Hav i ng control l ed th e exper i ment a l cond i t i ons
dur i ng each run the mean values must be measur ed . Uncer t a i n t y can
a r i se as a d i f f er enc e between th e measure d va l ue o f a paramet er and t he
unknown true value . The uncer t a i nt y can be reduced by r ep l i cate
determ i nat ions , b u t measurement s are suspe c t i b l e to sys t ema t i c errors
wh ich r ep l icate determ i nat ions wi l l not d i scern. Syst ema t i c error
c annot be eas i l y quan t i f i e d an d can only be m i ni m i sed by ensur i ng that

the measur ement t echni ques used ar e val i d and accurat e . Sources of
sy stema t i c err or inc lude ; u nwan t ed e dge h eat t r ans f er , i ns t rument
E xper i mental Pro cedur e and Dat a C o l l e c t i o n 58

cal ibrat io n error s a n d i nhomogene i t ies i n th e phase change mat e r ia l du e


to the presence of a i r voids or thermocou p l e w i r e .

Cle land & Ear l e ( 1 98 4 a ) d i scuss th e d i f f eren t sources o f error common l y


encoun tered i n phase change exper i mentat ion and g i ve a chec k l i s t that
enab l e s the mos t p r ob ab l e s ourc es of d i f ferences between ex per iments
and pre d i c t ions other than the pred ict i o n method er r or to b e asse s s ed .
E x pe r imen tal dat a are most u se fu l i f systemat i c error i s negl i g i b l e

compared w i th random er r ors , and the l a t t e r are mi n im i sed .

5 .2 C HO IC E OF PHASE CH ANGE MAT E R I ALS

The mate r i a l s u sed for phase change e xper i mentat i on mu st me et t he


cond i t i o ns l i s ted in Tab le 3 .2 , be cheap and easy to use , g i ve
r epr oduc i b l e resu l t s and sh ou l d have acc urately k nown the rmal

propert i es . Food mat e r i als are rar ely homogeneou s , each samp l e c an
on l y be u sed e xpe r i mental l y on ce or t w i ce , an d they t e nd to ha ve a w i de
var iat i o n in compos i t ion . Co nse quent ly therma l property data ar e often
i mp r ec i se . For these reasons ana l ogu es ha ve common l y b een u s ed ( R i ede l

1 9 6 0 a , Lentz 1 9 6 1 , Fraz erhur s t et a l 1 9 7 2 , Geuz e et al 1 972 , Badar i


Na rayana & Kr i s hna Mur thy 1 9 75 , Alb i n e t a l 1 97 9 ) . " Kar l sruh e t est
substance" developed by R i edel ( 1 9 60 a ) i s p robab ly the mo s t succes sful
and w i de l y u se d ( Flem i ng 1 9 67 , Bonac ina & Com i n i 1 97 1 , Bonac i na &
Com i n i 1 9 7 2 , Bonac ina et al 1 9 7 4 , Com i n i et a l 1 9 7 4a , Gorenflo & Mer t z
1 975 , Cle land 1 9 77 , Hayakawa e t a l 1 983 b ) . Common l y k nown and r e f erred
to as " Ty lose " , it is a 2 3 % me thy l-cellulose gel . Ty lose i s eas i ly
mou l d ed i n t o d i f feren t shapes and i s homogen eou s on ce e qu i l ibr a t i on has
occurred . It can be use d repet i t i v ely for exper i ment s wi thou t
dete r i orat ion , i t s dens i t y does no t a l t e r s i gn i f i ca nt l y as pha se c hange
occur s , and it has well c haract er i se d therma l proper t ies s i m i lar to
tho se of many h igh mo i s tur e foods ( Cle land 1 977 ) . For these r e a sons
mos t of the ex per imental wor k was done w i th Ty lose . The par t i c ul ar
Ty lose u sed was MH 1 000 � a produc t marke ted i n New Zea l an d by Hoechst
New Zealand Limi t e d .

O n mix i ng of the powdere d mater i a l w i th wat e r , rap i d hydrat i o n of the


gel meant that i n i t ia l l y it was non-homogen eous and some a ir
entra i nment occurred i n the f inal mou l ds . The water content o f th e ge l
E xp e r i mental Proc edur e and Data Collec t i o n 59

e qu i l i b rat ed over a per iod of several day s lead i ng to a homog e neou s


ma t er i a l . Air pock ets w i th in the gel tended t o b e very s ma l l ( l ess
than 1 mm in d i ame t e r ) and e v en l y d i s t r i bu t e d . B y a v i s ua l exam inat io n
of the f i na l g e l , the voids volume was es t i mated as less than 0.5%,

wh i l st br eak i ng th e join of the Ty l ose w i th f l a t met a l and p l a s t i c

sur fac es showed an almo st comp l e t e and ev en contac t . The pos s ib l e


effects on heat t r ansfer of i mp er f e c t con tac t are d i sc u ssed in
Se c . 5 . 4 . 3 . I t wou ld be expected that other r esea chers usi ng Ty l ose
have had s i mi l ar probl e ms w i th a i r en tra i nme nt , y et no dev i at i ons f rom
the repor t e d thermal p roper t i es have b een no te d . Th erefo r e it was
cons i de r ed that the unc ert a inty in the rmal pr oper t i es for Tyl ose was

not s i gn i f i c antly increased by the presenc e of the a ir bub bl e s .

Er rors i n mea sur i ng comp onents led to a mo i s ture con t ent of 7 7 . 0 ± 0 . 2%


i n the gel . D i rect me asuremen t of th e mo i s tur e in the ge l a f t er
exper i menta l work showed that except whe r e l i q u i d i ngress occurred
dur i ng freez i ng or thaw i ng by l i qu i d immer s i on ( d i scus sed in l a t er
secti o ns for shap es wher e th i s was a prob l em ) , the f i na l mo i s tur e was
w i th i n 0 . 5 % of th e or i g inal v a l u e .

C l e land & Ear l e ( 1 9 8 4 a ) g i v e the thermal c onduc t i v i ty and ap parent


vo lume t r i c spec i f i c heat c apac i t y ver sus temper atur e dat a f or Ty l ose
based on th e dat a of R i edel ( 1 9 60 a ) and Com i n i et al ( 1 974a) . Other
vers ions o f the therma l proper t y dat a have b e en u se d . C l e l and & Ea r l e
( 1 9 7 9 b ) use d a " f lat t ened" v er s i on wh ich enab l e d sub stan t i a l sav i ng s i n
comput a t i on t imes for nume r i ca l pred i c t io n method s , at a s l i gh t cost i n
t erms of loss of accuracy as l arge t ime steps cou ld be use d w i thou t
" j ump i ng " o f the latent heat peak . C l e land et al ( 1 9 82 ) tested a
hypothe t i ca l vo lume tr i c spec i f i c heat capac i t y c urve wh i c h h ad some
superc oo l i ng effec t s i ncorporated i n t o i t . The Ty los e thermal property
dat a u se d i n Group I I ( nume r i ca l ) me thods in th e present wor k are g i v en
i n Tab l e 5.1 . Compar i so ns w i th the or i g inal data from R i edel ( 1 9 6 0 a )
and dat a for m i nced l e an bee f ar e shown i n F i g . 5 . 1 a n d 5 . 2 . Val ues i n
Tabl e 5 . 1 were c hosen t o b e an accur a t e r ep r esen tat i on of R i e del ' s
data , but th e shape o f the volumetr i c spec i f i c heat capac i t y c urve was
altered so that when i ncorporated i n to the f i n i t e d i f f er ence method and
f i n i t e e lemen t me thod pr ograms , " j ump i ng " of th e latent he a t p eak was
less l i k e ly and large t i me steps st i l l gave good heat balanc es . Most
E x pe r i mental Procedure and Dat a Co l l ect i on 60

s i mple freez i ng and thaw i ng t i me p r e d i c t ion me thods need values of the


therma l pr oper t y dat a for on l y th e par t s o f the overa l l t emp erature
r ange where th e change i n therma l proper t i es w i th temperature is s ma l l .

T he therma l p roper t y dat a for Tylose u se d by Cleland ( 1 977 ) were

adop ted for Group I ( s i mp l e ) methods . These are g iven in Tab l e 5 . 2 and
are cons i s te nt w i th Tab l e 5 . 1 .

To show whether expe r i mental results obtaine d usi ng Tylose were t y p i ca l


o f the freez i ng and thaw i ng of real foods , e xper i men ts wer e a l so
conduc ted wi th mi nced lean bee f . Mu l t i p le fat and mo i s tu r e
det ermi nat i o ns g a v e th e compos i t i o n of the m i nced lean bee f as 74 .9±
1 . 9 % water and 3 . 1 ± 1 . 1 % fat . Mea t as a mat e r i a l i s less homogeneou s
than Ty lo se , so publ i she d therma l p roper t y data va r y s i gn i f i cant l y and
ar e not cons i s tent w i th each other ( Mor ley 1 9 7 2 ) . T he therma l p rop erty
dat a u se d we re de r i ved f rom R i edel ( 1 957 ) , Mor l ey ( 1 9 72 ) and f rom the
comp os i t i on factors o f Com i n i et al ( 1 9 7 4 a ) by C l e land & Ear l e ( 1 9 82a ) .
These are shown i n Tab les 5 . 1 and 5 . 2 and F ig s . 5 . 1 and 5 . 2 . T hey are
probab ly less prec i s e than the Tylose dat a .

5.3 T EM P ER ATUR E ME ASUR EM EN T AND C ON T ROL

Al l t empe rature me asuremen t s wer e made wi th 24 St anda rd W ir e Gauge


( SWG ) copper / cons t antan th ermocou ples ( 0 . 5 mm wire d i ame t e r , 0 . 9 mm
d iame t e r i nc l ud i ng p l as t i c i nsul a t i o n ) connect ed to e i the r a 12 po i nt

Taylor Ins truments "Mu l t i-Scan Recorder " p oten t i ome ter operat i ng on a

60 second print c yc l e or a 12 po int Hon eywe l l - Br o wn r ecord i ng


pot ent iome ter opera t i ng on a 1 00 seco nd p r i n t c yc le w i th an o p t i o nal
i ntermi t t en t 1 5 m inut e de l ay between cyc l es . Bo th of these mach ines
wer e ca l i b ra t e d w i th the abov e thermocouple wire t o w i th in 0 . 3 °C in the
r ange -50 °C to 5 0 °C .

Pr ior to thaw i ng or freez i ng the ob j ects of Ty lose or m i nced l ea n beef


wer e k ep t in temper atur e cont r o l l e d r ooms for l on g enou gh t o att a i n a
uni form temp eratu r e throu ghou t . To r educe the var i a t i o n in i n i tial
t emperature dur i ng the t i me fr om r emoval o f obj ec t s fr om the con s tant
t emperature areas t o the start of the f r eez i ng or thawi ng proc es s , the
obj ects were i ns u l a t ed . Excep t for th e s l ab shapes where sheets o f
0 . 05 m th i ck po lystyrene foam boar d wer e u sed , t h e obj ec t s wer e w r apped
Exper i mental Procedur e and Data Co llect ion 61

Tabl e 5 . 1 Thermal Property Data U sed In Catcula t i ons By Numer i ca l


Methods 1

T k T c H2

( oc ) (W m
-1 oc - 1 ) ( oc ) (MJ m - 3 o c - 1 ) ( MJ m - 3 )

Ty lose

- 40 . o 1 . 67 - 40 . 0 1 . 88 0.0
- 30 . 0 1 .6 7 -30 . 0 1 . 92 19 .0
-20 . 0 1 . 66 - 20 . 0 1 .95 �
8.3
-1 5 . 0 1 .6 4 -1 8 . 0 2 . 00 2.3
-1 0. 0 1 .63 -1 6 . 0 2 . 20 46 . 5
-9 . 0 1 . 61 -1 4 . 0 2. 0 � 51 .0
-8 . 0 1 . 60 -1 2 . 0 2. 0 56 . 1
-7 . 0 1 . 58 -1 0 .0 �
. 70 62 . 6
-6.0 1 .56 �
- .0 . 20 66 . 6
-5 . 0 1 .5 2 0 5 . 00 71 .2

-

-4 . 0 1 . 46 - .o 5 . 90 6.6
-3 . 0 1 . 35 l
- .0 7 . 20 3 .2
-2 . 5
-2. 0
1 . 28
1 . 18

- .0 1 1 00
1 7 . 00
• 92 .
1 06 .

- .0
-1 . 5 1 . 04 -3 . 0 25 . 00 1 27 . g
-1 .0 0 . 82 -2 .5 u 3 . 00 1 41 .
-0 . 8 0 .66 -2 . 0 5 . 00 1 61 3 •

-0 . 7 0 .55 -1 .5 70 . 0 0 1 90 . 1
-0.6 0 . 49 -1 . 1 00 . 00 207 . 1
0.0
20 . 0
0 . 49
0.56
-0 .
-0.5
6 1 00 . 00
1 9 .93
2 1 .1
2 J§
.1
40. 0 0 .6 2 0.0 3 .71 28 .o
1 00 . 0 0 .62 1 00 . 0 3 .7 1 660 . 0

M i nced Lean Beef

- 40 . 0 1 .58 - 40 . 0 1 . 89 0.0
-24. 0
-1 2 . 0
1 .5 � -30 . 0 1 .91 1 .o �
1 .4 -25 . 0 2.02 2 .8
-8 . 0 1 .40 - 20 . 0 2 .7 0 40 . 6
-4.0 1 . 28 -1 5 . 0 �
.58 56 . 3
-2.0
-1 . 0
0 . 98
0 . 48
-1 0 . 0 . 55 �
6 .7
-8 . 0 5 . 26 6 .5
o.o
1 00 . o
0 .49
0 .50
-6 . 0 7 .71 9 .4 �
-5 . 0 1 0 . 40 1 0 .5
-4 .0 1 5 .7 0 1 21 .5
-3.0 3 1 . 20 1 45 . 0
-2 . 0 53 . 1 0 1a .1
-1 . 5 73 .70 21 . 8 A
-1 . 3 25 5 . 00 251 . 7
-1 . 2 25 5 . 00 2 77 . 2
-1 . 0 3 . 65
3 . 65
� 03. 1
1 00 . 0 71 . 7

1 L i near interpolat i o n was use d .


2 Datum : H • o . o at -� . o oc .
Exper i mental Proc edur e and Dat a Co llect ion 62

Tab l e 5 . 2 Thermal Pr oper ty Dat a U sed In Calculat ions By Simple Formu l ae

Property Ty los e Minced Lean Beef

ks ( W m - 1 oc - 1 ) 1 .65 1 . 55
kl ( W m - 1 oc - 1 ) 0 . 55 0 . 50
CS ( MJ m -3 oc - 1 ) 1 . 90 1 .90
c
l ( MJ m -3 oc - 1 ) 3.71 3.65
L ( MJ m -3 ) 209 . 0 20 9 . 0
tJi (MJ m- 3 ) 226 . 0 23 0 . 0
Tif ( oc ) -0 . 6 -1 . 0
T i fave ( oc ) -2 . 1 -2 . 5
k ave ( W m - 1 oc - 1 ) 1 . 2f
1 ot•

1 �H entha lpy d i ffer ence between 0 °C and - l 0 °C , k ave


= =average
the rma l conduc t i v i t y dur i ng the pha se change pr ocess , T ifave mean =

freez i ng or thaw i ng temperatur e .


Ex peri m ent al Proc edure and Dat a C o l l ec t i o n 63

?
1.5


· · · · · · · •·
·· · ·
· ·· •
· · ·•
··
··
·· ·.

>-
··
··

-
·

·s;

.
.

:;: ·.
.

(J •
• Legend
-dc .


. A
0 .
.

u B
.


• .. . . . . . . . . •

.

.
E

0.5 ;;L---I'ftr.=

..c:.
I-

0 4-----�----�--��-
-20 -15 -10 -5 0 5 10
Temperature (•c)

Fig. 5. 1 Th e r ma l Conduc t i v i t y Da t a For T y l o se ( A ) a nd M i nc e d L e a n

Bee f ( B ) .

1,000


� Legend
� 100 A.
>-
a

1
···-·-·-

c
...............

I
u
;
u
,_.._______
cl

-20 -15 - 10 -5 0 5 10
Ter11*'oh.r• (•c)

F i g. 5 . 2 A pparent Volu met ric S pec ific Heat Capac it y Data Fo r T ylos e

( A , C ) and Minc ed Lean B eef ( B ) .

A , B - dat a from Tab le 5 . 1 , C - data from R i e de l ( 1 9 60 a) .


Exper i mental Procedur e and Data Col l ec t i on 64

i n insulat ion ma ter i a l and transported i n containers l ined w i th 0 .02 m


th i ck po lystyren e foam . Though the de l ay s we re as long as 1 0 m i nu tes ,
usi ng th is method the max i mum measured d i fferenc e in the ini tial
temperature of the obj ects from the mean value was found to b e 2 . 0 °C .
For s l ab s , b ec ause th e i nsulat ion cou l d be app l i ed and hel d more
d irec tly ont o th e sur faces , the temperature var iations were only ±0 . 5 °C
ab out the mean value .

For var i a t i ons i n temper atur e of up to 1 0 °C on e i th er side of a mean


i n i t i a l temperatur e the d i f feren ce in pr ed i ct e d freez i ng t i me for slab s
was found to be less than 0 . 5 % ( Cl e land 1 97 7 ) . Wh ilst th is analy s i s
was carr i ed out us i ng f i n i t e d i fferenc e s i mulat ion very s i mi l ar r esul ts
wou ld be expected if an exper imenta l invest igat ion was carr i e d ou t , as
f ini te d i f ferenc es prov id e a suf fi c i en t l y ac curate s i mulat ion of the
fr eez i ng process . Errors of s i mi lar magni tude would be expected for
shapes other than s l ab s . For thaw i ng the error du e to i n i t ia l
temperature non-un i form i ty wou ld b e expected to be smaller because the
enthalpy change for a 1 °C var i a t ion in i n i t i a l temperatur e is a lower
percentage of the total enthalpy change , c ompared w i th that for a 1 °C
var i a t i on i n ini t ia l temperatur e in fr eez i ng . Th e need to u se a mean
i n i t i al temperature to r epresent a non-un i form i n i tial temperature
there fore intr oduced negl i g i b l e error .

The amb ient heat i ng or coo l i ng med ium was e i th er water or 29 % ca lc ium
chlor ide b r i ne he ld i n a 1 m3 i nsulated t ank ( F i g . 5 . 8 ) . The med ium
temperature was measur e d by two th ermocoupl es ; one pr ior to and one
aft er the ex per imental sec t io n . E xternal c ir c ulation of the med ium was
used for sl ab thaw i ng ex p er i men t s , in wh i ch c ase an add i t i onal
thermoco uple was locate d i n the return stream to the tank . In all
expe r i men ts there was no measur abl e d i f ferenc e in temperature b etween
the thermocouples b ec ause the c irculat ion rate of the med ium w i th in the
p i p i ng and ar ound the t an k was suff i c i en t l y h ig h .

The t ank t emperature was c ontrolled b y a RKC PN- 4 1 " Bl i nd Control ler"
act iv a t i ng bo th refr i g erat ion and heat i ng systems . Th e thermal
capaci t y o f the tank contents was l arge c ompared w i th the heat loss or
ga i n from the env ironmen t and from the obj ec t be i ng frozen or thawed .
Exper imental Procedure and Dat a Collec t i o n 65

Th i s mean t that the med ium t emperature wou l d change on l y slow l y wi th


t ime if l e ft uncontrolled dur i ng the exper i ment . The control system
was suff i c i en t l y prec i se that no var i at io n of the temperature dur i ng
exper iments was detec tab l e w i th the recorders u sed .

Tak i ng r ecor der ca l ibrat ion error i nto account , the overal l err or in
temp era ture measurement and control was est ima ted a t l ess than ±0 . 5 °C .

5.4 THAW ING OF S LABS

5 .4 . 1 The Equ i pment

One-d imens ional heat transfer in an infi n i te slab undergoi ng f r eez i ng


c an b e approxi mated by an i nsula ted f i n i te- siz ed slab i n a pl a te
freezer ( Cl e land 1 9 77 ) . A s i mi lar arrangement shown in Fig . 5 . 3 was
used for thaw i ng of s l abs . Tw o plates wer e heat ed i n ser ies by water
pumpe d at a rat e of approx i mate l y k g s - 1 fr om the we l l m i x ed
temperature control led tank ( F i g . 5 . 8 ) . Any th ick ness of slab could be
accommodated . Therma l cont act betwe en th e sl ab surface and th e pl ates
was k ep t the same for all exper iment al runs by comp res s i ng th e plates
wi th p r es sur e prov ided by 1 00 kg of weight ove r the p l a te a r ea of
0 . 1 6 m2 •

For each slab th i c k ness a mou ld was ma de by cutt i ng a 0 . 22 m d iameter


ho le i n the c en t r e of a 0 . 4 m x 0 . 4 m p ie ce o f po lystyrene foam b oard
of the approp r i at e th icknes s . Two she ets of alumin i um fo i l w i th one o f
b rown paper between wer e u sed t o cove r o n e s i de o f the mou l d , a nd then
it was f i l led w i th Ty lose or mi nced lean b ee f from the op pos i te s i d e
t ak i ng care t o avo i d a ir entra i nmen t . Copper /const ant an thermocouples
were i nserted , two at d i f ferent pos i t ions on each sur fac e , and thre e at
d iffere nt pos i t io ns at the c entre th i ck n es s of th e slab . All
thermocou ples were run through i sotherma l r eg io ns ( para l l e l t o the slab
surfac es ) . Jun c t ions were located near the m i dd l e of t he 0 . 22 m
d iame t er mater i a l sec t ion to m i n im i se errors due to h eat conduct ion
a l o ng the w ires and e dge heat transfer effects. Two l aye rs of
alumin ium fo i l w i th one o f brown paper between were then f i x ed t o the
upper fac e to g iv e a f i l led moul d as shown in F i g . 5 . 4 . The wrapp i ng
added a smal l h eat transfer res i st anc e but prevented dehydrat ion of the
Exper i ment a l Procedure and Data Col l ec t ion 66

sur face of the mater i a l . The brown paper , p laced betwe en the two
l ayer s of a luminium foi l so that i t rema ine d dry , prov ide d structura l
strength t o the wr ap p ing . The mate r i a l was f i l l ed sl ightly ab ov e the
level of th e po lystyrene i nsu l a t ion to ensur e the insulat ion wou l d not
h inder contact b etween the plates and the sur face of the s lab .

5 .4 . 2 Th i ckness Control and M easurement

Th e s l a b th i ckn es s was measure d a fter equ i l ib rat ion over se vera l day s ,
i n both the frozen and unfroz en s t ates . Pr ior to thawi ng ex per i ments
th e s l ab s were fr oz en in a p l a te freezer t o ensure a flat sur face f or
good contact w i th the thaw i ng plates , even wh en so l i d and i n f l ex ib l e ,
and to g i ve the requ i r ed th i ck n ess . The th i ck ness of the Tyl ose sl ab s
was found to vary by up to 0 .5 mm across each slab . There was no
d i ff eren ce between the frozen and un froz en measureme nts for Tyl ose as
the dens i ty d id not change apprec iab l y . Thou gh the dens i ty change for
m i nc e d l e an bee f is of the or der of 4% some o f the var i a t i on i n vo lume
caused the poly styrene surround ing the slab to compress and d id no t
affect th e th i ck n es s t o th i s ful l ex ten t . The va l ues tak e n in the
froz en and unfr ozen s t ates wer e found to d i ffer by less than 0 . 5 mm
from the mean for meat . As the s ma l l es t s l ab th i ckness w i th mi nced
lean b ee f was 0 .024 m the er ror assoc iated w i th th is measur emen t i s
l es s than 2 % . Becau se a l l thermo coupl es wer e p l aced t owards the centre
of the 0 . 22 m d iameter s lab to avo i d edge effec t s in he at transfer ,
more empha s i s was p u t on th i ckness measuremen t s i n th i s reg i on .

5.4.3 Measurement and Control o f Sur face Heat Tr ansfer Coef f i c ients

In the plate thawi ng system use d the res is tanc e to heat transfer
between the s l ab surface and th e heat i ng medium was smal l , there were
no s i g n i f icant rad iat ion or mass transfer effec t s , and so th e boundary
cond i t io n was adequately descr i b ed by the th ir d k i nd of boundary
cond i ti o n . T o alter the surfac e heat transfer coeff ic ient ( h) vary ing
number s (0 to 1 0 ) of 1 . 5 mm th i ck , black " I nsert ion" rubber she ets w ere
placed on e i th er side of the test slab betwe en the p lates shown i n
F i g . 5 .3 . D iffer en t surfac e hea t t r ansf er coef f i c i e nts were obt a i ned
by alter i ng the number of sheets of rubber .
Exper imental Procedur e and Dat a Co l l ect ion 67

Four methods were u sed to measur e h: (a) a f ini t e d if fe rence


approx imat ion method , (b) a heat balance method , ( c ) the Goodman plot
and ( d ) an ana l y t i c a l method . Deta i l s o f the f ir s t three o f these are
g iven b y Cl e land & Ear l e ( 1 9 7 6 a ) . Bec ause sur fac e temper ature is mos t
sensi t i ve t o changes i n th e sur face h eat transf er coe f fic i ent ( Comi n i
1 97 2 ) , these three methods are based on me asurement of slab surfac e
temperatur e as i t v ar i es w i th t ime . The f ourth method u sed the
ana l y t i c a l solu t ion for he at conduct ion subj ect to the th ir d bou ndary
cond i t io n , for constan t th erma l proper t i es wh ere there i s no change of
phase ( Cars law & Jaeger 1 959 ) . The proc edur e i nvolved calcul at i ng
va l ues of the sur fac e heat transfer coeff ic i en t that f i t ted
exp er imental values of the surface temperature as it chang ed w i th t i me .

The therma l d iffus i v i ty of frozen ma ter i a l ( as ) is h igh so the


prehea t i ng t ime prior to phase change oc curr i ng dur ing a thaw i ng run i s
too short to obta in accurate est ima t io n of h by two o f th e four
method s . Also th e maj or i ty o f the heat transfer dur ing thawi ng w i l l b e
wi th th e surfac e o f the s l ab thawed . In the short t i me the s l ab
sur fac e was st i l l froz en the r i g id nature of th e slab may have resulted
in i mper f ec t cont ac t wi th th e p l at e surfaces g iv i ng effec t i ve ly l ower
surface heat transfer coeff i c i ents . For these reasons , separate
exp er imen ts conduc ted at temperatures gre a t er than the in i t i a l freez i ng
point of Tylose were used to determi ne th e sur fac e heat transfer
coeff i c ient rather than r e ly ing on preheat i ng data. Abov e 0 °C th erma l
propert i es c an be cons i dered constant and a l l four methods c ou l d b e
appl ied equally wel l .

A total o f 3 3 runs u s ing a var i ety of Ty lose slab th i ck nesse s , and


amb i en t and i n i t ia l temperatures we r e u se d to det ermi ne h . I f each o f
the ru bber she ets was i ncomp ress i b l e and prov ided equal resi stanc e to
heat t r ansf er a l ine ar va r i a t io n of th e r es istanc e t o heat transf er
{ 1 /h ) wi th the number of rubber sheets should oc cur ( Cowe l l & Namor
1 97 4 ) . The maxi mu m number o f rubber sheets u se d was t en and the p l a te
to plate pressure was kep t cons tant at a l ow valu e , so i t is unl i k e ly
that s igni f i can t var i a t io n i n r es istan ce e i ther betwe en th e sheets or
due to the compres s io n of the rubber oc cur re d . The heat capac i ty of
the rubber i tsel f may a l so affect exp er i ments . By preheat i ng the
rubber on the plates the amou n t of heat transfer r equ ired t o change the
Exper i mental Procedure and Data Co l l ec t io n 68

ru bber temperature was kept min imal compared w i th that n eeded to heat
the actual s l ab mater i a l in b oth tha w i ng e xper i me nts and runs to
determine h. If the heat ing of the rubber was s i g n i f i c an t , the
est i mated h val ues wou l d be lowe r than tho se ac tua l l y occur r i ng and
wou ld appear to vary w i th t ime . The later trend was no t ap parent for
any of the methods so th i s effect was cons i de r ed negl ig ib l e .

Ther e was no s ign i f i c ant d i ffer enc es b etween values of h calculated


from th e four me thods at the 95% l eve l of conf i denc e . Also the
d i fference in values was i nsign i f i cant compar ing exper iment s wh ere th e
s l ab s were coo l ed w i th tho se where they wer e heate d . The re was no
corr e la t i o n of h w i th the slab th ickneses or the amb ient and i n it i a l
temperatur es that were u se d in these runs . Figure 5 . 5 g i ves the l east
squares regress ion l i n e relat i ng ex ternal h eat transf er res i s tanc e
( 1 /h ) to th e numb er of she ets o f rubb er . Var i at i on i n the ind i v i dua l l y
measured h values showe d that the error i n the average es t i mates of h
for par t i c ular numb ers of rubber sheets was ±2 . 2% .

Becau se o f the pr eparat ion of s l ab s by pre-freez i ng and th e u se of a


constant pressure , the var iat ion i n surface heat transfer coe f f ic ient
between r eg io ns on th e sur fac e of th e s l ab was co nsi der ed n egl i g ib l e .
Ev i denc e t o th is effec t was fou nd from rep l i c ate surface the rmoc ou p l es .
Var i a t io n i n recorded temperature was less than 1 . 0 °C i n any thaw i ng
run .

It i s poss i b l e that surface heat transfer coeff i c ients determi ned from
th e separat e ex per i ments are no t r ep r es entat i v e of h dur i ng thaw i ng
runs . However the heat transfer cond i ti ons dur i ng the separate
exper imen ts were s imi l ar to tho se dur i ng each thawi ng run , ex cep t for
the i n i t i a l stages i n the thawi ng runs where th e slab surfac e was st i ll
froz en . The s lab s were ini t ia l l y froz en i n a p l a te freeze r and the
rubber sheets mounted between the p lates and the slab i ts el f had some
f l ex ib i l it y . There fore the cont act between the plates and the s l ab
surface was considered cons tant throughout the thawi ng expe r i menta l
runs . Furthe r , pred i c t ions of surface temp erature by numer i cal methods
wer e as good i n i t i ally as they were later in the exper imental runs .
Th i s conf i rms that these est i mates of h were rep resentat i v e o f the
ac tual values throughout thawi ng . Us i ng a number of separate runs
Ex per i mental Proc edure a nd Data Co l l ec t io n 69

A
B

1'---- - ---1----1 .......


E F

F i g . 5 . 3 Schema t i c D i a gram of the Exper i me n tal Slab Thawing Equipme n t .


No t t o sca l e . A - hea t i ng p l at e s , B - rubber shee t s , C -
i n su lated test slab , D - th ermo c ou ples , E - cool i ng plates , F
- heat i ng elemen t s , G - c irculat io n pump , H - insu lat ed ,
ag i tated, temp eratu r e control led tank .
E xp er i mental Procedure and Da ta Co l l ec t i on 70

. .

Fi g . 5 . 4 Cons truc t i o n of Test Slab s .


( a ) a comp leted slab
( b ) a slab w i th i nsulat ion r emov ed.
E xper imental Proc edur e and Dat a Collec t io n 71

0.10

0.08

.......

3: 0.06

E
u
0

..c 0.04

0.02

0 2 4 6 8 10
No. of Rubber Sheets

F i g . 5 . 5 Pred i c t ion of Sur face Heat Transfer Coe f f ic ien ts For Slab
Thaw ing Exper imen ts .
Exper i menta l Procedure and Da ta Co l l ec t i on 72

meant tha t h was det ermi n ed accurat e ly , yet as i ndependen tly as


poss i b l e of the thaw i ng t ime ex per i me nts .

To confi rm that h v a lues es t i mated i n the separate expe r i ment s were


r ep resentat i v e of va lues dur i ng thaw i ng , some ana ly s i s dur i ng ac t ua l
thaw i ng runs was done us ing all the me thods d isc ussed prev iou sly .
Ther e was no d i ffer e nce in these resu l ts compared wi th th e sep ara te
exper iments .

For the thawi ng of th e mi nced lean beef slabs the sur face heat transfer
was assumed to be the same as for the Tyl ose s l ab s . Ther e was no
reason to suspect that h wou ld be d iff eren t beca use mos t of the
r es i stan ce to h eat transf er was suppl i ed by the ru bb er sh ee ts , the same
slab cons truc t ion was used and the we ight ing of th e slabs ensured good
therma l con tac t for both mat er ia ls .

5.4.4 Ana ly si s of Heat Tra nsfer i n Slab s

In a true one-d imens ional slab there i s no heat transfer i n the other
two d i mens ions . I n prac t i ce th i s i s i mposs i b l e t o ach i eve . To r educe
heat trans fer in the other two d irections three techn i ques were
emp l oyed :
( a ) the other two d i mens ions were mad e l arger than the slab th icknes s ,
wh i ch i s the cr i t ical d i mension b y a t leas t a f ac tor of 2 . 1 : 1
( b ) each s lab was surrounded on the edges by at l e ast 0 . 09m of
po lystyrene foam i nsu l at ion, and
( c ) the thermo co uples were placed near the centre of each 0 . 2 2m
d iame t er s l ab .
For the wor st case o f a l arge slab b e i ng thawe d slowly the effect of
edge e f f ects due to he at transfe r thr ough the i nsu l at i on ar e shown by
the analy s i s in App . B to decrease thawing t ime b y 4 . 6 % . Th is ana ly s is
took the most pess imi s t i c v ie w of the s i tuat io n . For a mor e t yp ic a l
case th e effect of edge heat transfer o n thawi ng t i mes was l ess than
1 .0%. The s l ow nature of thaw i ng means that edge heat t r ansf er c annot
be avo ided , so when data are interpreted i t must be r emember ed that
some exper i men t s wi th s l abs wer e not tru l y one - d i mens i ona l . I n onl y 6
out of 35 slab ex per i ments was the effec t on thaw i ng t i me s of h eat
t r ansf er through t he e dges , f ound to be mor e than 1 .0 % .
Exper imenta l Pr oc edur e and Dat a Co l l ec t ion 73

A i r entra inmen t a nd th e r esu ltan t l es s than p er f ec t therma l contac t


were ment ioned in Sec . 5 . 2 . The latter problem d id not occur i n
exper i menta l wor k w i th s l ab s bec au se th e we i gh t i ng of the p l a tes to
hold the s lab s t i ghtly in p lace ensur ed good contact . Pr e-fr eez ing the
s l ab s i n a p l a te fr eez er me an t that the pro tec t i v e l ayer of a luminium
fo i l was alwa ys t i ghtly and comp letely bound to the frozen mater i a l .
Because a ir vo i ds in the Ty l ose ge l w ere s ma l l and uni forml y
d i stributed the ir e ffec t i s only impor tant i f they a ffect the ov era l l
heat transfer behav i our of a s l ab . Go od pred i c t i on of e xp er i menta l
sur fac e and centre temperature prof i les wer e obtained by nume r i c a l
me thods for freez i ng ( Cleland & E ar l e 1 97 7 b ) and for thaw i ng
( Se c . 7 . 2 ) . These results ind icate that the rma l propert i es use d wer e
representat i v e of the Ty l ose mate r ia l inc lud i ng voids , thu s i mp ly i ng
that the ef fec t of the vo ids i s neg l i g ibly sma l l .

There are a lso error s consequ en t on th e th ermocouple placeme nt w i th i n


each slab . Al l thermoc ou ples were inserted paral l e l to the heat
transf er surface thr ough supp osed ly i so thermal reg i ons . There fore , the
only temperature grad ient for conduc t i o n along the th ermocou p l e wires
was due to heat transfer from edge ef fect s , wh i ch has been shown ab ove
to be genera lly sma l l . Also the w i r e cross -se ct ional ar ea i s less than
0 . 1 % of the area ava i l ab l e for heat t ransfer due to s i de e f f ec t s , so
ov erall the effec t of th e co nduc t io n along the w i r es was ins ign i f ican t .

O f the to tal a r ea norma l t o th e d irec t ion of heat f l ow at each surfac e ,


and a t th e centre , only 0 . 5 % to 1 . 0 % was covered by th ermocoupl e wire
and the i r p l as t i c i nsulat i on . As th e gre ate r r es ist an ce to heat
transfer prov ided by the plas t i c i nsulat ion ag a inst that of Ty lose only
appl i es for th i s l i mi t ed are a the thermocoupl e s did no t af fec t heat
f low s ign i f icant ly . The th ermocou ple l eads oc c up ied less than 0 . 0 5% of
the gel vo lume so th e heat c apac i ty of th e slab was no t s ign i f ic an t ly
affected by the i r p resenc e .

Temperature me asurements were t ak en a t th e two mo st impor tan t places ;


th e surf a ce a nd the cen t r e of th e s l ab s . The former i s s ens i t i ve to
changes i n the ex ternal heat transfer cond i t i ons and so was used in
est imat io n of the surfac e heat trans f er c oe f f i c i en t . Th e l a t t er was
needed to measure the thawi ng t ime . Acc ur ate p l acemen t of the
Exper i me ntal Procedure and Data Col l ec t i on 74

thermocouples at th e sur fac e and cen tre wh i l e mak ing the slabs i s
d if f i c ul t and m i gra t io n of the thermoc ou ples w i th in the gel aft e r
const r uc t ion is pos s ib l e . For these reasons more tha n o ne thermocouple
was u se d to me asur e each tempera tur e .

I t was i mposs i b l e for a the rmocoup l e to be ab ove the slab sur face
because of the plate confi gurat ion . Prov ided hea t transfer is
homogen eous acros s the sur face , the fastes t hea t i ng thermocoupl e w i ll
be the best est imate of the sur fac e temperature . A true est imate of
surface temperature wou l d be g i ven b y a th ermocoupl e j u s t touch i n g the
layer of alumi n i um fo i l , so any sl ower heat ing the rmo couple is l i k e ly
to b e below th e surface . Howeve r i n thaw i ng th ere is no de f in i t i ve
cr i t erion to d i s t i ngu ish between d i ffer enc es in temperature due to
i nhomogenei t y and tho se due to the rmocouple placemen t , such as the
su percoo l i ng behav iour observed in fr eez i ng by Cl e land ( 1 9 7 7 ) . Th e
only o th er cri t er io n that can be used is wh ether the surf ace
temp erature prof i les flat ten not icab ly ( plateau ) in the l atent heat
temperature range or no t . Becau se th e sur face i s th e f irst reg i on to
thaw and the temp erature change c an not be slowed by the presenc e of
lat ent heat add i t i on in r eg io ns ou ts i de th e surface, any the rmocouples
r i ght at the sur fac e will tend to "sl ide" through th e l atent heat
temperature r ange rathe r than show the ty p ical p l ateau effec t
char acter ised b y mo r e cen tral pos i t io ns . Unfortunat e ly , th e speed that
th e surface tempe rature pas ses throu gh th i s temperatur e range is
dependent on the ov era l l rate of heat transfer . As amb i ent
t emp eratures c l ose to the thaw i ng t emperatures wer e often u se d i t was
d if f i c ult to f i nd a c lear d is t i nc t ion b etwe en good and bad th ermo couple
pos i t i on i ng . For other amb i en t cond i t i ons some d i f fere nt ia t i on cou l d
be made . F i gure 5 . 6 shows typ i c al tempera ture prof iles f o r d i fferen t
thermocoup les near the sur fac e . Curve A i s typ ic a l of a we ll p l a ced
surfac e thermoco up l e . Its prof i l e i s bar ely a ffected by latent heat
release in the thaw i ng temperature range , but has a p l a teau at a
temper ature between th is and the amb ient temperature . Curve B i s for a
l ess prec i se l y pos i t ioned thermocoupl e . The temperature pro f il e
flattens c ons i derably due t o l atent heat before r i s ing becau se the
thermocoupl e i s be low the surfac e . Curve C i s for the very bad case
wher e the thermocou ple is we l l below the surface and the rate of
t emperature change w i th t i me is slowed marked l y by the absorp t i on of
E xper imental Pr ocedur e and Data Co l lec t io n 75

l atent h eat .

C l early , th is c r i t e r ion does no t comp l et e ly d is t i ngu ish between good


and b ad sur face thermocoupl e plac ements , espec ially where overa l l rates
of heat transfer ar e low . After de let ing obv iou s badly placed
thermocoupl es i t was found that ind iv i d ua l thermocoup l es gave
measurements of the , surface temperature of the slab that var i ed by less
than 1 . o oc ar ound th e ir mean valu e . Ty p i c a l ly one or two ou t of four
sur face thermocouples wer e found to be poorly placed and were no t use d .
After exper imenta t i on placemen t was ch ecked b y d i sma nt l i ng th e s l ab s .
The obse rv ed th ermocoup l e pos i t ions c ompar ed well w i th thos e suggested
by th e above ana ly s i s and in a l l cases at least one thermoc oup l e was
wel l placed .

Hea t transfer cond it ions on b oth faces of the slab were as equ i va lent
as poss i b l e so th e geometr i c and thermodynami c c entres shou l d co i nc i de .
Any thermocou ple no t exac t ly at the cen tre w i l l change in temper ature
more qu i ck ly than one at the cen tre . Prov i ded he at transf er is
homogeneous across the slab sur fac e and w i th in the slab , the l as t
thermocoup l e to reach 0 °C w i l l g i v e the best est i mate of the slab
c entre temperatur e and the most accurate est imat e of the thaw i ng t ime .

When food thaws most of the l atent heat is ab sorbed at temperatur es


j u s t below the i n i t ia l fr eez i ng temperature ( T i f ) . An ab rup t cha nge i n
therma l prope r t i es oc cur a t T if ' r esu l t i ng in a su dden incr ease i n
temp erature . Cons ider a s l i ghtly mi splaced centre thermo co up l e .
Thaw i ng will no t be comp le t ed through ou t the s l ab wh en th i s
thermo co upl e reaches T i f • Hence a rap id change in th e temperature w i l l
not occ ur a s heat transfer i s s t i l l r equired to supply latent heat t o
more c entral r eg ions . Th e measur ed temperature tends to rema in
'

r elat i ve ly co nstant at some temperature j ust abov e T i f depend ing on how


far i t i s from the centr e . When the l ast r eg ions of the obj e c t thaw ,
a l l the t e mperatures · ( e v en at the surface thoug h th e e ff ec t there i s
less appar en t ) undergo a sec ond increase i n r ate of temperature change
wi th t ime. Th i s is b e cau se h eat transf er is now r equ ired only for
sens i b l e hea t ing . The further from the c entre the smal l er th is second
"br eak" i n th e tempe rature p rof i le wi l l be . A s i mi l ar e f fe c t occurs i n
free z i ng but is less not i cable b ec ause the change f r om latent he at t o
Exper i ment a l Proc edure and Data Col l ec t i on 76

sensi b l e heat abso r p t i o n i s l ess d is t i nct than that i n thawing . James


et a l ( 1 976 ) used th i s pr inc i p l e t o pred i c t pha se change t i mes by on l y
measur ing surface temperatur e . I t can also be used to imp rov e the
est i mate of tha w i ng t i mes from the rmoco upl es not placed ac curate l y at
the c entre . Where the slowest heat i ng th ermo coup l e d id not " b reak "
away from th e l at en t heat temperature range and immed i ate l y head
asymp to t i c al ly for the amb i ent temperature , th is method was use d to
est i mate th e tha w i ng t ime . Tangen t ia l ext ensions to the temp erature
pro f i le i mmed iately prior to and immed iate ly after the second "b reak "
po in t we re drawn and the thaw i ng t i me t aken at the t ime ind ica ted by
the ir i n tersec t ion as shown in Fi g . 5 . 7 .

The err or i n est ima t i ng th e thaw i ng t i me by th i s b reakpo in t ana ly s i s


was inves t i gated in two way s . Firstly , u s ing exper imental d a t a where
one we l l p l a ced c en t r e thermocoup l e could be compar ed w i th another
cent r e thermocou p l e in the same slab . The t i me pred ic ted usi ng th is
ana l y s i s on the second thermocoup l e was compa red wi th the t i me for the
well p lac ed thermocou ple to reach 0 °C . In all cases stud i ed the
d i fferenc e was l ess than 0 . 5 % of th e tota l t i me a l though the ex ac t
d i stance the se co nd thermoc ou ple was from the actua l thermo d ynamic
c entre was unknown ( d i sman t l i ng o f the s l ab s showed that a ll centre
thermoc oup les were w i th in 5% of the total th ick ness from the geometr i c
c entre o f th e s l a b ) . Secondl y , f i ni t e e l ement and f in i te d i f f erence
me thods w ere use d so the e ffect of the er ror in the pos i tion of th e
thermocoup l e s c ou l d b e s tud ied quan t i t a t i ve ly . The same e f f ec t was
shown. Us i ng th is method the pred icted freez ing t i me s for a
thermocoup l e d i splaced 1 0% o f th e d istance from the c entre to t he
sur fac e were w i th in 0 . 5 % of the true values of 0 °C at the thermo d ynam i c
centre for the w i d e r ange of cond i t io ns s tud ied . The e f f ec t of centre
temperature thermo co up l e placement was therefore not s i g n i f icant in
es t ima tes o f thaw i ng t i me but mus t be cons i d ered when compar i ng
pred icted temperture prof i les . Only experimental runs that had good
c entre a nd sur face th ermocoup l e pos i t io n i ng we re used for tempe ratur e
pro f i l e comp ar i son .

For obj ec ts w i th mor e than one d imensio n th is same break p o i n t analy s i s


w i l l a l so a pp l y . In fac t , because for these obj ec ts l es s th an 5% o f
the volume i s enclosed w i th in a d is t ance 5% from th e thermodynamic
Exper i mental Pr oc edur e and Data Collection 77

10

.
.....
. ..
...
u .
. .
0
...__
0 t------:>,__ ___ / .. -- T; ,
-:#----� -
,.__ •

.. ..
.

I /
I !

I
=
-10
Legend
I A

I B

c
-20 �--�----r---�--.--.

nme
Fig. 5 . 6 Ty p i c a l T e mp e r a t u r e Pr o f i l e s For Th e r mo co u p l e s Pos i t io n e d A t

Or N e a r t h e Sur f a c e o f a Th a w i n g S l a b . A - at t h e s u r f ac e , B

- 0 . 0 025 m b e l ow the sur fac e , C - 0 . 005 m b e l ow t h e s u r f ac e .


-2
D 0. 1 m, h = 50 . 0 i-i m oc - l , T = 1 3 . 0 °C , T = - 2 0 . 0 °C ,
a in

T
if
= -0 . 6 °C .
3 .
·

.
.
.

.

.
.

2 :

1/
..·,
u t2
.
� .
«»
3 ....
.
- tl
0 •

L.. ..
• •
••

«» ..

!
••
0

Legend
-1 A
8
c
. . . . . . ...••

-2 �----r---��--�--.---T---r-
10 10.5 11 11.5 12 12.5 13 13.5
Time (hrs}
Fig . 5 .7 Break point Analy s i s to Es t i ma t e Thaw ing Times From
Thermocoup l es Not Posi t ioned Exac t l y at the Th ermod y namic
Centre . A - a th ermocouple at the thermodynam i c cent r e , B - a
thermocoup l e 0 . 0 05 m from the thermodynami c cent r e , C - a
· thermo c ouple 0 . 0 1 m from the thermodynamic centr e . D • 0 . 1 m,
h - 50 . 0 w m- 2 o c- 1 , T a • 1 3 .0 °C , T i n . - 20 . 0 °C , T i f 2 - 0 . 6 °C ,
t • est i ma ted thaw i ng t i me from A , t 1 • est i mated thaw i ng
t ime from B , t 2 • est i ma ted thaw i ng t i me from C .
E xp er ime ntal Procedure and Da ta Co l l ec t ion 78

centre the bre ak effec t w i l l be magn i f ied . Cl e land ( 1 977 ) showed that
a th ermo coup l e p l aced 1 0% of a spheres rad ius i n error , t he measur ed
freez ing t i me wou ld be 0 . 6 % in error e ven i f the ab ov e analy si s was not
used . Any err or in p l acemen t of th e c en t r e thermoc oupl e was h i gh l y
damp ed i n i t s effec t on fr eez ing t i me , and d id not s i gn i f icantly
i nc r ease the overa l l err or . For s l ab thaw i ng , th e err or for a 1 0%
d isp lac ement was found to i ncr ease as ( T a - T i f ) increased but by us ing
the b reakpoint me thod of analy s i s was only 0 . 1 % for T a = 1 3 °C . The
break point analy si s was therefore used for all shapes inv e s t i gated
where c entre the rmo coup les we re shown to be not acc urate ly plac ed at
the th ermodynamic centre .

An est i mate of the exper i mental er ror ar i s ing from imprec ise control of
experi mental cond i t ions for th e s l ab experi men t s was determin ed from
rep l i c ate thaw ing runs . As far as pos s ible , th e repl icates wer e
i ndep en den t of each oth er inc lud i ng the construc t i on of ne w s l ab s for
each run . After norma l i sa t ion of th e rep l i cates the variab i l i ty showed
that th e e xper imenta l error ar i s i ng fr om poor cont ro l corresponded to
95 % co nfidenc e bounds of ±5 . 2% . The total exper imen tal error wi l l be
h igher than th i s due to unc er tainty in measur ement of the e x per imental
cond i t ions and sour ces o f syst emat i c er ror .

5 .5 TH AW ING OF I N F I N ITE CY LIND ER S

5 .5 . 1 The Equ i pment

To obta in un i form heat t ransfer cond i t ions at th e sur face for shap es
other than s l ab s a l iqu i d immers ion thaw i ng sys tem was u sed . Liqu i d
immer s ion g ives much less var iat ion of the surface heat transfer
coe f f i c i en t and amb i e nt temperatur e ar ound an obj ec t than th e a ir
thawi ng sy stem wh ich i s also commonly use d . A schema t i c d iagram o f th e
l iqu i d i mmer s i on t ank i s shown in Fig . 5 . 8 .

The i mp e l l e r c irculated the thawing med ium, water , around the


2 . 1 m x 0 .7 m x 0 . 7 m i nsulated tank. The l iqu i d passed over
r e fr igera t ion plates and then heat i ng e leme n ts before flowing through
the expe r i men ta l sec t i on ( 0 . 8 m x 0 .5 m x 0 .7 m ) a nd back to the
i mpeller . Two mesh screens ( ap er ature size 1 mm x 1 mm) prov ided
Exper ime n tal Pr oc edur e and Dat a Col l ec t i o n 79

suff ic ien t head los s to equa l i se any un evenness i n the l i qu i d f l ow .


The c irculat ion was su f f ic iently h igh that the t ank acts as i f i t was a
p er f ec t ly m i x ed sy s tem. The temperatur e was control l ed to w i th in
±0 . 3 °C i n the range 0 °C to 5 0 °C for thaw ing ex per iments ( Se c . 5 . 3 ) .

Thoug h the var i a t ion i n surface h eat trans f er cond i t ions shou l d be
sma l l in the l i qu id system , the test samp les were osc i l lated throughou t
the exp er imen ts to ensure tha t obj ec t or i enta t i on w i th in the flow
stream d id no t af fec t un i formi ty of sur fac e cond it ions . The obj e c t s
were tur ned throu gh an ang le of 300 ° every 30 se cond s . The sa mp l e
osc i l lator i s shown in Figs . 5 . 9 , 5 . 1 0 a n d 5 . 1 2 .

In order to ap prox imate i n f in i t e ly l ong cyl inder s , five d i f fe rent


leng ths o f cy l indr ica l p i pe 0 . 45 m to 0 . 50 m long were u s ed . Stee l
p i pes of 0 . 1 m and 0 . 1 5 m nominal d iameter and polyv iny l chlor i de ( P VC )
plas t i c p i pes of 0 . 05 m , 0.1 m and 0. 15 m nom ina l d iame t er gave
d i f ferent heat transfer res i s tanc es due to the d i fferent wa l l ma ter i a ls
and th i cknesse s .

The c y l i nders were pack ed w i th Ty lose and thermo couples int roduc ed at
the c entre and surfac e pos i t ions from e ach end throug h i s othe rmal
reg ions . Two th ermocouples at the c entre and four at the surface were
used for each cyl inde r . To redu ce end he at trans f er e ffec ts ,
poly styren e foam insulat i o n cap s of at least 0 . 05 m th ick ness were pu t
on the end of each cyl i nder and th e c y l inder clamped in th e samp l e
holder for osc il lat i o n a s shown in Fi g s . 5 . 9 , 5.10 and 5. 1 1 . The
thermo coupl e ent r y points and the po lystyrene cap s wer e sea l ed to
prevent l i quid contac t w i th the Ty los e .

5 . 5. 2 D iame ter Cont rol a nd Measurement

Ther e was no s igni f ica nt d i ffer ence in d iameter measuremen ts b etween


the fro z en and unfrozen s t ates for the c y l i nders . Any dens i ty chang e
that occurred has a long i tud ina l e ffect tha t was ab sorbed by
compr ess i on o f the po lystyrene foam cap s . The d iame te r was measure d to
an accuracy of ±0 .5 mm .
Exper i mental Procedur e and Dat a Col l ec t i on 80

r
,.

� �

.,
� +

A B


� � tI C
/'
, D
,,
,,
E �
,, '•
,,
/
'• F
'• G
.,
'• ' '

Fig. 5 . 8 S c h e ma t i c D i a g r a m o f t h e L i q u i d Imme r s i o n Ta n k .

A p p r o x i ma t e sca l e 1 : 1 5 . A - h e a t i n g e l e me n t s , B - c oo l i n g

p l d t es , C - i mp e l lor for c ircu !.lt ion , D - ba f f l e plate , E -

mesh screens , F - e x p e r i me n t a l s ec t io n , G - i nsula ted t an k .

A
B


F

F i g . 5 . 9 Schema t i c D i a gram of the System U sed To Ho ld and Osc i l l a t e the


Inf ini t e C y l inder s and Two- D i mens ional Irregular Shapes i n the
L i q u i d Immers ion Tank . Approx ima t e sca l e 1 : 5 .
A - attachmen t to the sa mp l e o sc i l lator , B - s t and attached t o
t h e imme r s io n tank , C - c lamp to ho l d the test obj e c t , D -
i nsul at e d c ap s , E - the t e s t obj ec t , F - thermocoupl es .
Exper i m e n t a l Pr o c e d ur e a n d Data C o l l e c t i on 81

Fig. 5. 1 0 The Sa mp l e Osc i l l a t or a n d Inf ini te C y l i nd e r Tha w i n g E q u i p m e n t

Used in the Liqu id I mmer s i o n Ta n k .

D E

Fig. 5. 1 1 Schema t i c D i a g ram Sh ow i ng the Arrangemen t o f t h e Po l y s t y r e n e

Foam C a p s a n d Thermo c ou p l es Leads For I n f i n i t e C y l i n d er

E x pe r i men t s . No t t o s ca l e .

A - p o l y s t y r e n e f o a m cap s , B - t h e r mo c o u p l e l e a d s , c - p i pe

wal l , D - s u r fac e t h e r mo c oup l e s , E - c en t r e the r mo co u p l es .


Exper i mental Procedure and Data Col l ec t ion 82

5 .5 .3 Measurement and Control o f Sur fac e He at Transfer Coeff i c ients

The d i f ferenc e in cyl i nder wa l l mat e r i a l a nd th ickness .save some


var i at ion in th e sur face hea t t r ansf er coeff ic i en t . To g i ve greater
var iat ion of h , and to reduce th e surface heat transfer coe f f i c ient to
values more common ly encountered in a ir th aw i ng sys t ems , zer o to four
sheets of rubber were glued to the ou ts ide of each cyl i nder . The
rubbe r u sed was th e same as tha t u sed to a l t er h in th e s l a b thaw i ng
exper imen ts . The same l i n ear r e lat ionsh ip be tween heat trans fer
r es i st ance and th e numbe r o f rubber shee ts u sed for s l ab s cou l d not be
ap p l i e d because of the c urvature of th e c y l inder surface ar ea and the
need for glue to permanan t l y attach the rubber . Consequen t l y separate
est imates of th e surface heat t rans fer coeff i c ient were ma de for each
comb ina t ion of p i pe mater i a l , wal l th i cknes s and number of rubb er
sheets .

For the same reasons as those ou t l ined in Sec . 5 . 4 . 3 for slab s ,


separate heat i ng or coo l i ng e xper imen t s wi th ou t pha se change were
necessary to accurately es t i mate h. To determine h for the ten
d i f fer e nt comb i nat ions out l i ned ab ove , a tot a l of 29 runs were mad e .
I n i t i a lly the thr ee me thods use d by C l e land & Earle ( 1 97 9a ) were
app l ied . Simi lar to the method s u s ed to de t ermi ne h for s l ab s , the se
methods a l l depended on the me asur eme n t of surface temperature as it
var ied w i th t i me . I t was found tha t th e resu l t s were h i gh l y v ar i ab l e
and i ncons i s ten t . Th i s was thou ght to b e due t h e e ffect of i mperfect
cont ac t of the Tylo se to th e i n s i d e sur fac e of th e cy l inde r wa l l s .
Imp erfect contact was no t a major p rob l em i n fr eez ing ex per i me n ts u s ing
th e same equ ipmen t ( Cl e l a nd 1 977 ) as the s l i gh t ex pans ion of the Tyl ose
and contrac t ion of the p ip e wal l ma t e r i a l dur i ng fr eez ing f i l led any
vo i d s . In thaw i ng , s l ight shr ink i ng of the Ty l ose and e xpans ion o f the
p ip e wall mat e r i a l o cc ur r e d and th i s c reated voids .

D i sman t l i ng o f the c y l i nder s a fter exper i menta t i on sh owed that thermal


contact was b e t ter than 80% . Where perfect contact did not oc cur a
smal l gap was p r es en t between the p ipe wal l and the Tylose. The se
voids were randomly d i s t r ibuted and i l l-def in ed in s ize so it was
i mposs i b l e to e va l u a te the i r i nd i v i du a l e f fec t d ir ec t l y . The b e s t way
Exper imen tal Pr ocedur e and Dat a Collec t io n 83

to accoun t for them was to u se an average h value for the who l e


surface . The mode of heat transfer across th ese voids would be natural
convec t ion . I f he at t r ansfer acros s th e gap s was by conduct i on on l y
then a 0 .5 mm void wou ld lead to a s i g n i f icant res i s tanc e of
oc
-1
0 . 02 5 m2 w • The actual heat transfer res is tanc e due to natural
convec t i on wou l d b e l ess than th i s , b u t how much l ess can no t b e easi ly
pred icted .

Ind iv idual sur face thermo co uples wou ld be affected by the ir prox im i t y
to the se voids . Henc e the va lue of h es t imat ed f rom sur face
temp erature me asurements would be affec ted and may no t be
r ep r es entat i v e of the t ru e average value . Becau se there was no
c r i terion to chos e wh ich thermocou ple mo s t c los ely resemb led the
av erage sur fac e cond i t ions th e me thods based on sur face temperature
cou ld not be rel i e d on to g i v e good e s t i mates of h .

A me thod that cons i dered th e c entre temperature was the re fore u sed . By
usi ng t h e cen tre temperature the effect of a ny var i a t io n in heat
transfer c ond i t ions at th e sur fac e was e f fec t i ve l y averaged and the h
value was more representat i ve o f the true valu e . Loca l i sed variat ion
in h is no t i mport ant i f i t i s re lat i ve l y smal l in magn i tude and i s
randomly d is tr i b uted ove r the surface. Th is was probably the case w i th
the sur face voids and i mper fe c t cont ac t .

The me thod use d t o est i ma t e the surface heat t ransfer coef f ic ient was
the h ea t p ene t ra t ion method ( Ar c e & Swea t 1 980 ) . For l ong t i mes the
analy t ical solution for heat transfer w i th cons tant the rmal propert ies
and the th ird k ind of b ounda ry cond i t ion ( Ca rslaw & Jaeg er 1 9 59 ) is
close ly ap prox imated by tak ing only the first term in the ser ies
so lu t ion . The sur fac e he a t t rans f er coeff ic i en t c an b e c a lculate d f rom
the slope of a plot of ln Ye versu s t i me ( Charm 1 9 6 3 , Bak er & Cha rm
1 9 6 9 , Kopelman e t a l 1 9 7 0 ) . An advan tage of th is me thod i s that i t is
i ndependent o f the exac t pos i t i on of th e thermoc oup l e wi th i n the
cyl inder . I f th e c entre th ermocou ple i s not p rec i s ely loc ated on the
c en tral ax i s , th i s does no t a f f ec t the est ima t i on o f h b y th i s metho d .

A n i n d i c at i on that the presence of v o ids was affe c t i ng surface


temperature read i ng s was that the read i ng c lo sest to the a mb i en t med ium
Exp er i me n tal Pr oc edure and Dat a Collec t i o n 84

temperature d i d no t c orrespond t o th e same thermocoupl e throug hou t the


run . The best est i ma te of th e centre t emperature was t ak en as th e
ther mo c ouple record i ng the temperature th at was slowes t to change
towar ds th e amb ient t emperatur e . The two centra l the rmocou ples t ended
to show temperature p rof i les cons is ten t l y d i f f erent fr om each other ,
ind icat ing that inhomogene i ty was l argely a surface problem and that
any e f fects of non-un i formi ty at the sur face were h igh l y damp ed i n
the ir e ffec t o n heat transfer near the c entre .

The rubb er shee t s gave a cons t an t add i t iona l res istan ce to heat
transfer over the whole cyl i nder sur face excep t for the j o i n t n ec essa ry
down th e leng th of the cyl inder . The rubb er was cut ac curate l y so that
th i s jo int was alway s less than 0 . 5 mm in w idth . For the sma l le s t
cyl i nder th is af fected 0 .3% of the sur face ar ea , wh i ch is
ins ign if icant . Wher e mu lt iple layer s of rubber were glued on , the
j o ints were no t over l apped so the change in sur fac e cond i t ions cau sed
by the jo ints wer e not compounded at one se ct ion of th e cyl inder
sur fac e . Th e contact adhes i ve used gave extreme l y good cont ac t b e tween
rubber layers and the p i pe walls w i th no a ir spaces so heat transfer
thr oug h the rubber l ayer s was uni form .

Thaw ing ex per i ments were carr i e d ou t wi th the amb ient me d ium
t emperatur e in the range : 5 . 0 °C to 4 5 . 0 °C . Al though the l iq u id
ve loc i t y ar ound the tank wa s constan t , some var i a t i on of the transport
proper t i e s of water and ther e fore the f i lm h eat transfer coeff ic ient
may occur wi th temperatur e . I n the tempera tur e range us e d th e changes
in l i qu i d propert i e s were small , and the b ulk of th e res i s tanc e to hea t
t r ansfer was i n th e p i pe wa ll and sheets o f rubb er anyw ay . Hen ce there
was no stat i s t i c al l y s i gn i f ic ant correlat ion of h w i th amb ient
temp erature .

Centre temperature prof i l e s change l ess than surface t emperatures for a


g iven change i n h so ther e was a loss o f measuremen t sens i t iv i t y . The
result was that ind i v idual h values d if fe red by up to 20% from the mean
v a lu e i n se t s of repl i ca t e runs for the same cyl i nde r and r ubb er
th i c k nesse s . The unc erta i n ty in th e mea n est i ma tes of h were th erefore
ab ou t ± 7 . 0 % at the 9 5 % level o f conf idenc e .
Exper i me n ta l Pr ocedur e and Dat a Col lec t io n 85

5 .5 . 4 Ana lys i s of Heat Transfer in Inf ini te Cyl ind er s

Some o f the prob lems enc oun te red i n cy l inder expe r i ments we re s i m ilar
to those found for slab s . The leng th s of the test cyl inders were
a l way s great er than the d iame t er b y a fac tor of at l east thr ee , and
insula t i ng caps were used to m i n i mi s e the effect of heat transfer along
the leng th of the cy l inders . Us i ng a s i mi l ar ana ly s i s to that for
slabs ( Sec . 5 . 4 . 4 and App . B ) the end-e ffe c t heat transfer was found to
b e neg l ig i b l y small ( 0 . 2% decrease in thaw i ng t i me ) for the worst
si tuat ion use d in thawing runs . The test cyl inders th erefore
appr ox i ma t ed inf in i te cyl inder s very close l y .

Apart from the prob lems at the sur fac e , air voids that oc curred in the
mater ia l were no la rge r or more common than th ose found for slab s , so
the ir e ffec t wou ld not be ex pec ted to be any mo r e s ign i f ican t than the
effec t of the in ternal vo ids for slab s . Th i s was d i scussed in
Se c . 5 . 4 . 4 . Vo ids a t the surfa c e of th e cy l i nder s and imperfect
contac t predom inan t ly a f f ec ted the sur face h eat transfer a nd were
accounted for by the me thod d i sc ussed in Se c . 5 . 5 . 3 . A consequenc e was
that the geome t r i c and thermo dynami c c en t r e of the cy l inder may not
have co inc ided. As surface voidage was ev enly d is t r i b uted across the
cyl inder sur fac e s , averag i ng e f fects mean tha t th e c entral ax i s of the
cyl inders shou ld coinc ide closely w i th th e thermodynami c c en tre . It
w i l l b e shown below that e v en i f the thermodynamic c en t r e was m i sp laced
s l i ghtly , the thaw i ng t i me i s only min ima lly a lt ered .

Err ors due to inhomogen e i ty created by the pr esen ce of th ermo coup l e


w i r e s and heat c onduc tion a lo ng th ese wires were of s i m i la r magn i t ud e
to tho se fo und for slab s , and wer e s i mi lar ly ins igni f ic an t . Er rors in
plac emen t of centre th ermocouples were assessed b y th e same me th ods
used for s l ab s . Cen t r e ther moc oup l e p l acemen t was mor e d i f f i c u l t than
for slab s so a greater var i a t i o n in t emperature was ob served between
thermocouples . As shown in Sec . 5 .4 .4 qu i t e large err ors in
thermo coup l e plac eme n t c an o ccur w i thout s i gn i f icantly affec t i ng
pred i c t io n o f thawi ng t i me or c entre tempera ture pro f iles . P l a cement
of thermoco uples n ear the thermodynamic c entre was the refore eas i ly
ass es sed and d i d not s igni f icant ly increase the ove ra ll error . The
breakpoin t analy s i s ou t l i n e d in Sec . 5 . 4 . 4 was u se d t o increase the
Expe r i mental Pr ocedure and Data Col l ec t ion 86

prec is ion of thawi ng t i me pred ict ions whe r e centre thermoco uples were
poor l y pos i t ioned . For reasons d i scus sed in Sec . 5 . 5 . 3 the fastes t
chang ing ind i v idua l surface temperature measurement d id not n ec es sar i l y
r ep r es en t the ave r age h ea t transfer cond i t ions , and slower changes
could be due to placement below the sur fac e as well as lower h values
i n par t i c u l ar reg ions . Hen ce app l i cat ions of th e techn i q ues of
Se c . 5 . 5 . 3 p robab ly l e d to ex per i menta l ly measured surface t emperatures
nearer to the amb ien t temperature than th ose r epresen t i ng th e av erage
heat transfer cond i tions . Pr edic ted surface temperatures usi ng the
average h va lue would therefore be expec ted to l i e further fr om the
amb ient me d ium t emp erature than the me asured data.

The effec t of the heat capac ity of th e rubb er insulat i on and a l so of


the p i pe wall ma ter ial was greater than that d i sc ussed in Sec . 5 . 4 . 3
for s l ab s , because the rubber and p i pe wal l mate r ia l c ou l d not be
prehea ted to the amb ient thawing temperature due to it be ing
pe rman en t l y att ache d to or part of each cyl i nder . The heat t ak en up by
the rub ber and wa l l ma ter i a l was less than 3 .0% of the total h eat
t r ansfer to e ach cyl i nder . Fur th er , th e es t ima te of the sur fac e heat
transfer coeff ic ient to the Ty lose sur fac e was determined in such a way
that on l y heat t r ansferred throug h the insu l at i ng l ayer s was t aken i nto
account . Hence the ma in e ffect of heat tak en up by the rubber and p i p e
wal l mat er ia l wou l d b e to d i stor t t o some ext e nt th e me asur ed Ty l ose
surfac e temp erature early in the thawing proc es s when the rubber and
wal l temperatures wer e chang i ng most .

No r ep l i cate runs were co nduc ted for c yl i nder thawing due to


l i mi t a t ions in t i me , so d irect measuremen t of th e exper i mental er ror
was no t made . The error s for cy l i nder thawi ng were o f the same order
of ma gn i tude as in the s l ab expe r i men t s , excep t for measuremen t and
control of th e sur face heat transfer coeff ic ient . The ex per i me n ta l
error w a s there fore expected to b e sl ight l y gre a t er tha n for s l ab s .
Exper i me n tal Pr ocedur e and Dat a Coll e c t ion 87

5 .6 THAW I NG O F SPHERES

5 .6 . 1 The Equ ipment

Ho l l ow me tal b a l ls were used to mo del spher i c al geome t ry . The th ree


s iz e s u sed were approx ima te l y 0 . 05 m, 0 . 1 m and 0 . 1 25 m i n d iame ter .
The me tals wer e sta inless steel , c opper and mi ld steel respect i ve ly .
Each of the sphe res was made fr om t wo hemi spheres for ease of
constru c t i on . Each hemi sphere was pack ed w i th Ty lose and two sur face
thermo coup les wer e i nser t ed a long i so thermal pa ths around the
c ircumferenc e . Th e two halves were then welded or soldered together
wi th a s i ng l e the r mo couple in th e c entre po si t ion . Th e d i s rup t i on to
the surface of each sphere cau sed by the joint between the two
hemi s pheres a f f ected 3 . 5% to 4 .0% of the sur fac e are a . However the
j o int s were of s i mi lar therma l res i s tanc e to other parts of the
sur fac e . The d i s rup t ions and prob lems cau sed by voids and i mper f ec t
contact are only i mport ant bec ause of th e ir e f fect on the heat t ransfer
b ehav i our of the spheres an d are d i scus sed in Sec s . 5 . 6 . 3 and 5 . 6 . 4 .

A bolt was soldered or welded to the surface of each bal l to enab le a


connect ion to be made to the sa mp l e ho l der and o sc i l lator . B ecau se
both the bolt and the spheres were me tal and only 0 . 04 % to 0 . 2 % of th e
tota l sur fac e are a was affected, the e f f ec t on h eat trans f er was
con s i d ered neg l i g ib l e . A f in i shed sphere and the samp l e holder are
shown in F i g . 5 . 1 2 .

The same imme r s i on thaw i ng tank ( F i g . 5 . 8 ) used for infinite c y l inder


exper i men ts was used for spheres . Osc i l la t ions o f the samp l e s was u sed
for th e reasons ou t l ined in Se c . 5 . 5 . 1 .

Sea l i ng t he ou ts i d e of each sphe r e a l ong the j o i nt between the


hemi spheres to prevent l iquid i ngress was a problem, espec i a l l y wh ere
the thermocoupl e w i res en tered the meta l she l l s . Sl ight water i ngress
l e d to changes in the Ty lose near the l eak , and often r esulted i n
dens i t y changes and h e nce spl i t t i ng of the sphe res on fr eez i ng . To
prevent th i s prob le m and to g i v e var i a t ions i n surface heat t ransfer
coeff ic i ents ( Se c . 5 . 6 . 3 ) l ayer s o f s i l i c one rubber ( Dow Corn i ng 31 10
RTV ) were use d to coat the spheres after cons tru c t ion . The rubber was
E x p e r i me n t a l Pr oc edure and Dat a Co l l e c t i o n 88

Fig . 5. 1 2 The Samp l e Osc i l l a t o r a n d S ph e r e Shapes U se d in th e L i qu i d

Imme r s io n Ta nk .
Exper imental Pr ocedur e and Dat a Collec t io n 89

app l i e d i n a l i qu i d form and th e spher es were r ot a t ed unt i l the rubber


had cured to m i n i m i s e variat ions i n the rubber th ickness . I n sp i te of
the i s prec au t io n someome loca l ised var i a t io n of rubb er th i ckness d id
occur . By ap ply ing mu l t i ple lay ers of rubber th is variat i o n was no t
large and was random l y d i str i buted . Th e e f f ec t on the sur face heat
transfer coeff ic ient is d iscussed in Sec . 5 . 6 . 3 .

5 .6.2 D iame ter Control and Measurement

The cons t ruc t ion of th e spheres in two ha lves , the f l ex ibi l it y of the
ho llow me tal she lls use d and the d i stor t ion in shap es due to water
i ngress and the co nsequ e nt densi ty chang e , mea nt that the spheres were
not cons tant in d iame t er . Some variat ion in d iameter occurred w i th
bo th p os i t ion ar ound the spher e and whe ther the spheres wer e in th e
frozen or un frozen sta t e . Measured v a r i a t i o ns in d iame ter ab ou t the
mean value were ± 1 . 0 mm ( 1 . 8 % ) for the sma llest sphere , ±3 . 0 mm (2 .7%)
for the me d ium s i z ed sphere and ± 2 . 0 mm ( 1 . 6 % ) for th e large sphere.
Th e sphe re vo lume was measur ed by d i splac eme nt of l iqu i d and conf i rmed
that the average values were accurate .

5 .6 .3 Measuremen t and Control of Surface Heat Transfer Coeff ic i en ts

The met a l bal ls had l ow r es i st anc e to hea t t rans f er , so mos t of the


change in h was due to the lay ers of s i l i co ne rubber use d to coat and
sea l the sur face . Some var i a t ion was ach i ev e d by vary i ng the numb er of
lay er s o f rubber a dde d .

The same prob lems w i th therma l contac t and voids tha t were observed
wi th inf i n i t e cyl inder s ( Sec . 5 . 5 . 3 ) occur r ed for spheres. In
add it ion , exac t cont rol of the rubber c oa t i ng th ickness was not
poss i b l e and irregu l ar i t i es in the met a l j o in t bet ween th e s phere
halves a ffec ted heat t ransfer . Therefore some var i a t i o n i n the surface
heat t r an sfe r coeff ic i en t ove r th e sur f ace was exp ected . Es t imates of
h based on sur fac e t emperature me asureme n t a nd t h e methods for r a d i a l
geomet ry o f Cleland & Earle ( 1 97 9 a ) d id no t p rov i d e con s i s tent and
r e l iab l e r esult s . For these reasons , the heat penetrat ion method
( Charm 1 9 63 ) was u sed to me asure h us i ng the c en t r e temperature
pro f i l e s fr om independent heat i ng and c ool i ng exp e r i ments tha t involved
E xper i mental Procedur e and Data Co l l ec t i on 90

no change of phase . As d isc ussed for the i n f in i te c yl inder geome try in


Sec . 5 . 5 . 3 th i s me thod averages any random surf ace v ar i at ion i n h . The
only poss ible sourc e of sy stema t i c variat ion was the j o i n t bet we en
hemi spher es . The j o i n i ng mater ia l was metal l i c in nature and there fore
had n egl i g ible add i t io nal r e s i s tance to h eat transfer compared w i th th e
o the r par ts o f the met a l sur face .

Howev er , there were more sourc es of error i n measur i ng and control l i ng


th e sur face h eat transfe r coe f f ic i en t for sph eres than for inf in i te
cyl inde r s . A total o f 3 2 runs were used t o est i ma te the t e n d i ffer ent
h va l ues u sed in thaw i ng experi ments . Var i at ions of measur ed h va l ues
between rel icate runs i nd icated that the measurement unc erta i n ty in the
mean values of h was ab ou t ± 1 1 . 0% .

5 .6 . 4 Ana lys i s of Heat Transfe r in Sphe res

For sphe r ical geomet r y there a r e of c our se no probl ems w i th end he at


transfer effec t s . There we re error s due to inhomogenei t ies on the
spher e sur f a ce at the j o i nt between hemi spheres , unevenn ess in the
rubber coat ing , at tachme n t of the bolt to al low connec t io n to the
samp l e o sc i l la tor and th ose d ue to a ir vo ids and i mperfe c t cont ac t .
The ex tent to wh ich these factor s a ffec ted temperatures r ecorded by the
sur face and cen t r e thermo coupl es cou l d not be eas i l y assessed but would
be s i m i la r to that for i n f i n i te c yl inders . The major e ffect was on
measured sur fac e temperatures wh i ch was d i scussed in Sec . 5 . 6 . 3 . The
pos i t io n of the actual th ermod ynamic cen tre was probab ly not great ly
a f fec ted by non-homogenei ty of sur face cond i t ions if they were randomly
d istr i b ut e d . Therefore th e measured centre temperature and thawi ng
t ime wer e onl y sl i gh t l y decreased i n accuracy by the se effec ts .

For spheres , the thermodynamic centre and th e geome t r i c c en tre co i nc ide


at a s i ng l e p o in t so that on l y one thermocouple cou l d be placed at th i s
pos i t ion . Er ror s due to th e i nhomogen e i ty created by presence of
thermocouple w ir e and placemen t o f thermocoup l e s wer e s i mi la r to tho se
found for i n f in i te c y l i nder s , and s i mi larly had ins ign i f ican t e ffec t on
p red i c t io n of tha w i ng t i mes . However , in spheres it was not poss ible
for the centre th ermo co uple to be inse r t e d along isoth ermal r eg ions .
Hea t conduc t ion wi l l onl y b e s ign if i cant a l o ng th e w i r e , no t a l ong the
E xper imental Pr oc edur e and Da t a Co l lec t i o n 91

insu l a t ion . As the cross -sec t ional area o f the w ir e ( 3 .6 x 1 0 -7 m2 ) i s


only 0 . 0 0 4 % o f the total surface area o f the smaller sphere and th e
thermal conduc t iv i t y of th e w ire i s abou t 200 t ime s tha t of Ty l ose ,
then up to 0 . 8 % of the to tal heat flow may occur through the wires .
Overa l l , the e ffec t of heat conduc t ion a long the c entra l th ermoc oup l e
w i r e o n thawi ng t i me was cons idered to be small .

As se s smen t of sur face and c entre th ermo c oupl e plac eme nts and est i ma t i on
of thaw ing t i me from exper i men tal runs was ma de in th e same way as that
u s ed for inf ini t e cyl inder s ( Sec . 5 . 5 . 4 ) . The coa t i ng of rubbe r u sed
on the spheres was far less than that used for i n f i n i te cyl inder s , so
th e ef fec t of th e heat t aken up by the rubbe r and me t a l she l l s on the
over all hea t trans fer was ins ign i f ican t . The volume of water that was
abl e to pene tra t e each sphere ar ound th e th ermo c ouple leads was
ex tr eme ly sma l l and only affec ted the Ty lose in the imme d iate v ic in i ty
o f th e l eak . The maj or result was some loca l i sed expans ion when the
Ty lose was fr oz en . Th is affec ted the d imens ions and phy s i c al i n teg r i t y
of the j o i nt s . An avera ge d iame t er valu e for th e s phe res based o n b oth
froz en and un fr ozen s t ates was use d . Sph eres in wh ich th e j o i n ts
b ecame v i s ib l y damaged wer e remade be fore fur ther u se .

Ov era l l , ex per i men tal er ror for spheres was o f a s imi lar ma gn i tu de to
that for thaw i ng o f inf ini te cyl inder s .

5 .7 TH AW ING OF REC TANGU LAR BR ICKS

5 .7 . 1 The Equ ipme n t

The l i q u id imme r s i o n thaw i ng tank was use d . Both me tal and


polypropy l en e box es wer e const ruc ted ( Fi g . 5 . 1 3 ) . All box es h ad a bo l t
a t t ached t o the l i d t o hold them onto the sample osc i llator i n th e same
way as that us e d for spher es . The fac e of e ach box fur thes t f rom the
geome t r i c centre was chosen to b e the l id . Th i s surfac e has the l eas t
e ffec t o n heat t ransfer i n the rectangular b r i ck and the bo l t a f f ec ted
less than 0 . 1 % of th e total surface are a , so i n all cases th e effect of
the bo l t was insign i f i cant .
E xp e ri mental Pr ocedur e and Dat a Co l lec t i o n 92

Cons t ruc t ion of the me t a l box e s , from shee ts of 1 . 2 mm th i ck mild


stee l , r equ ired some ove r lapp ing of the me tal to al low at tachment o f
the l i d b y screw s as shown in Fi g . 5 . 1 4 d . For me t a l boxes the ma j or
resi stanc e to heat transfer at the sur fac es i s th e l i quid f i lm h eat
t ransfer coeff ic i en t . Also the over l app i ng o c curred for less than 25%
of th e lid ar ea . Therefore th e d oubl e me tal lay ers at any ov er lap s
r esul ted i n n eg l ig ible add i t ional heat t rans f er res istan ce . By we ld i ng
and bend ing corner s , all edges of the me tal boxes ( F i g . 5 . 1 4 b ) c losely
app rox i ma t ed a sha r p corner ed b r ick shape ( Fig . 5 . 1 4 a ) excep t for the
l id .

The other boxes were ma de from polyp ropylene plas t i c sh eet ( 5 . 0 mm


th i c k ) wi th a l l the j o in ts screwe d tog ethe r ( F i g . 5 . 1 4 c ) . Th e presence
of th e screws altered the heat trans fer res i s tanc e of less than 0 . 4 % of
th e sur fa ce a r ea depend i ng on the box. Th e ef f ec t of th i s was
insign i f i c ant and the cor ners approx imated c losely to the id eal t y pe of
j o in t shown in Fi g . 5 . 1 4 a .

Thermoc ouples were pos i t ioned a t the geome t r i c c entre o f each brick and
at four to s i x d i f feren t p l ac es on the sur fac e . Whe re po s s ib l e ,
surface th ermo cou ples were pos i t ioned a t the c en tre of th e b r ick faces .

5 .7 . 2 D i mensi onal Measure men t and Cont ro l

The d imens ions of each rect a ngular b r i ck wer e measur ed in bo th the


froz en and unfrozen s tate s . Wa ter i ngress a round the thermoco uple
leads en try po in t i n the l i d o f some boxes resul ted in some l ocal
densi ty changes and henc e d i me ns ional var i ab i l i ty . Dens ity change
dur i ng phase change was a s l ight prob l em for the min ced lean bee f
b r i ck s . Al low i ng for th ese factor s , plus b ox cons t ru c t i o n and
measurement imprec is ion, th e d i mens ions were accurate to ±1 . 0 mm for
a l l the Tyl ose and m i nced l ean bee f rect angular br i ck s . The e f f ec t of
wat e r ingress on heat transf e r is d iscussed in Sec . 5 . 7 . 4 .
E xper iment al Pr oc edure and Dat a Collect ion 93

5 .7 .3 Measurement and Control of Surface He at Transfer Coe ff ic i e nt s

The metal a nd po lypr opylene b r i ck s gave two d i f fer ent surface heat
transfer coeff ic ient values . To further alter the sur fac e heat
transfer coe f f ic i en t , rubb er shee t s were g lu ed to e ach fac e of the
rectangular b r i c k s in a s i mi la r manner to that use d for i n f in it e
cyl inder s . On e l ay er o f rubbe r was p u t on th e metal boxes and two
lay ers on the poly propylene box es . I n th is way fou r d i fferen t v alues
of h were obt a ined for the rect angul ar b r i ck geometry .

I n a rectangular b r ick shaped obj e c t heat transfer at any po i n t on th e


sur fac e c an be treated as one-d i mens iona l un t i l the t i me that th e h eat
penetra t ion fronts from one or both of the o ther two d i mens ions reaches
th e point . Th i s means tha t the methods us ed for the measurement of h
i n slab s can be ap p l i e d at some pla c es on the surfaces of recta ngular
br icks for shor t t i mes . In Sec . 5 . 4 . 3 it wa s shown that th e four
me thods u sed for slab s we re equally accurate . Th erefore the mo st
conven i en t o f these me thods to use for shor t t imes , tha t ba sed on
Goodman ' s heat balanc e integral t echn ique ( Cl e land & Earle 1 9 7 6 a ) , was
the f ir s t me thod cho sen to de term i ne h .

The analy t ical p roduc t solu t i o n for he at conduc t ion i n three d ime ns io ns
wi th the th ird k ind of boundary cond i t i on and constant therma l
prope r t i e s ( Newrnan 1 9 36 ) was th e se cond me thod use d . A s i mi lar
comp ut er p rogram to that us e d to es t i ma t e h from sur face temperature
pro f i les for radial geome t ry ( C l e la nd & Earle 1 9 7 9 a ) was u se d . The
pr ogram took v a l ues of t i me , sur fac e tempera tur e and p o s i t iona l
coord inates for the pos i t i o n of th e th ermoco uple where th e me asurements
were mad e , and u s ed an i tera t i v e loop t o pred i c t th e sur fac e heat
transfer coeffic ient needed to a t tain tha t temperature , at th at t i me ,
and i n that pos i t i on .

Both method s are o nly appl i cable for t imes b e fore the onse t of phase
change . As th i s t i me i s v er y shor t for thaw i ng expe r i men t s , separate
hea t i ng and cool i ng runs were c ar r i e d ou t w i th each rectangular bri ck
i n the unfroz en s t at e . There was no s i gn i f i can t var ia t ion i n es t i mated
h values betwe en bricks of the same wal l ma ter i a ls , betwe en analy se s
us i ng d i f fe rent thermo c ouples i n each b r i ck , or as amb i ent med ium
E xpe r i me ntal Pr ocedure and Da t a Co llect ion 94

tempera ture change d . The d i sagreemen t of the two methods was


sta t i s t i c a lly ins ign i f icant at the 95% level of s ign i f icanc e .
Consequ en t ly , avera ge values o f h for e ach of th e four comb ina t i ons of
wall mater i a l and numb er of rubber sh eets use d , were calculated from
the r epl icate es t i ma t io ns .

Er ror s due to incons is tenc ies in heat transfer on each of the surfaces
of each rec tangular b r i ck wer e ma de n egl i g i b l e by us i ng th e l iqu i d
immer s ion tank , osc i l lat i ng the samples , and us ing th e rubber sheets to
a l te r th e sur f ace hea t transf er res ist anc e . The mean pred ic ted h va l ue
for poly propylene boxes had 9 5% conf idenc e bounds o f ± 1 . 5 % ( no ru bber )
and ±2 . 1 % ( wi th two she ets of rub ber ) . For the me tal boxes th e resul t s
wer e l ess acc urate th an those for th e po lypropylene box es becau se the
sur fac e heat tr ans fer coeff ic ients were larger and cons equen tly th e
t i me for each heat ing or cool i ng expe r i men t was sh ort er than for the
poly propy lene b r i ck s . The values of h were th erefore based on less
data . The 95% co nf idenc e bounds for th e mean values of ±7 . 5 % ( no
rubber ) and ±4 . 4% ( w i th one rubber she et ) for th e me tal b rick s , reflect
th i s t r end .

As for the slab ex per i ment s , th e h values for runs w i th minc ed lean
beef wer e assumed to be th e same as tho se measur ed for the Ty l ose
brick s . M i nc e lean beef undergoes s i g n i f icant dens i ty changes dur i ng
phase change and has a f ibrous mo i s t n a ture comp ared wi th th e dry
gel-l i k e nature of Ty los e . Th erefore a ir v o i d s w i th in the minced lean
b ee f were less l ike l y and i t was poss i b l e that the contac t r e s i s tanc e
was d i f fe r en t for the two mater ials . Henc e the assump t i o n o f equal h
v a lues may introduc e some extra uncer ta in ty for the m in ce l ean beef
exp e r i me n ts .

5 .7 . 4 Analy s i s of He at Transfer i n Rectangular Br ic k s

Error s i n exper i me n ts f o r rec tang ular b r i ck s aros e from s im i l a r sources


to tho se for o th er shapes . Vo ids a nd i ngress o f water oc curred to the
same ex tent as in spheres and wou ld n o t be ex pected to b e a ny more
s ign i f icant for rec tangu l ar br i ck shapes than the rad ial c a se .
S i m i la rly , any inhomogene i ty caused by the presence of thermoco uple
w ir e s w ou l d have a n egl ig ible e f f ec t .
E xper ime n tal Pr ocedur e and Dat a Co llec t io n 95

Pr ob lems had been found wi th imper fec t ther ma l conta c t for inf in i t e
cyl i nders and spheres . Si mi lar behav iour was no t ob se rv e d f o r th e
r ect angul ar br i cks . Therefore sur fac e tempe ratures and surface h eat
transfer could be analy sed in a si mi lar manner to that u sed for slab s
wi th out add i t ional er ror .

Su itable isothermal re gions for thermocou pl e l e ad locat ion do no t oc c ur


i n r ectangu l ar b r ic k shap ed ob j e c t s . Thermo coup l es were intr oduced
through reg ions as near to i soth erma l as poss ible to m i n i mi s e
conduc t ion a lo ng th e w ir es . He a t f l uxes a lo ng th e w i res wer e the re fore
s i m i l ar to those that oc curred for spheres . The volume of the brick
shapes u sed were much la rger than th e sma l l est sphere so the e f f ec t of
hea t transfer along wir es was ins ign i f ican t .

Error s i n placement o f thermocouples were ass es sed in the same w ay as


for oth er shapes . Because sur fac e hea t t ransfe r cond i t ions wer e
equ iva l en t on eac h fac e , the geome tr i c and the rmod ynam i c centres o f the
recta ngular b r ic k s c o i nc ide at a s i ngle point and , s i mi larly to a
spher e , on ly on e thermocoupl e c ou l d be plac ed there . As shown for
other sha p es (Se c . 5.4 .4) the error in plac emen t of the centre
th ermocoup l e cou l d be ea s i l y ass es sed b y th e br eakpo in t ana lys i s and
do es not s ign i f ican tly a ffect the ac curacy w i th wh ich thawing t i mes
c ou l d be measured .

As was the c ase for spheres , th e volume of the rubber coa t i ng and box
wal l mat er ial for the r ectangular br i ck s was suff ic ien t ly smal l
compared w i th the volume o f the Ty lose b r i c k i tse l f t o mak e the effect
o f the heat ab sorbed by the rubber and wal l ma t er ia l un impor tant .

For thawi ng of rec t angular b r i c k s the ov erall e x pe r i mental error was


smaller tha n for tha w i ng of inf ini t e cy l inder s and spheres ( bec au se h
cou ld be measured mo r e acc urately ) , and abou t the same as that for
s l ab s .
Exper imental Pr ocedur e and Da ta Co l le c t i o n 96

b
F i g . 5 . 1 3 Ty p ic a l R ectangular Br i ck Sh apes .
( a ) poly propylene box
( b ) meta l b ox wi th on e l ay er of rubbe r
( c) poly p rop y l e n e box on the sa mp l e osc i l lator ( p . 9 7 ) .
Exp er imen ta l Pr ocedur e and Data C o l l ec t i o n 97

-
-

Fi g . 5 . 1 3 Ty p i cal R ect angular Br i ck Shapes .


( a ) po lypropy l en e box ( p . 96 )
( b ) me t al box w i t h on e layer of rubber ( p . 9 6 )
( c ) po lypropylene box o n t h e sa mp l e o sc i l l a tor

c d

F i g . 5 . 1 4 Schema t i c D i agrams o f Box Cor ner Ty pes .


( a ) i dea l corn e r , ( b ) met a l b o x corner
( c ) poly prop y l e n e box cor ner , ( d ) metal box cor n er w i t h l i d .
E xpe r i mental Procedure and Data Col l ec t io n 98

5 .8 H EAT TR ANSFER IN TWO- DIM ENSIONA L IRREGULAR SHAP ES

5.8.1 The Equ ipme n t

A n i n f i n i t e numb er o f two-d i mens ional ir regular shapes are poss ible .


The ma in concerns in cho osing exp er i mental shapes were to b e ab l e to
( a ) a ccurately control and desc r i b e th e shap e , ( b ) control and me asure
th e sur face heat trans f er condi t ions and ( c ) preven t heat transf er i n
the th ird d i mens io n . It was cons idered that these cou ld be bes t
ach ieved by us i ng e xperi mental techn iques s i mi l ar to those u sed for
thaw i ng of i n f i n i t e cy l i nder s ( Sec . 5 . 5 ) .

Th e shapes were cons truc ted from 0 . 45 m to 0 . 50 m l e ng ths of 0 . 1 m and


0 . 1 5 m nominal d iame ter polyv iny l chlor i d e ( P VC ) plas t i c p ipes . These
p i pes wer e heat ed un t i l th e wa l ls became so ft and pl i ab l e . They were
then mou lded leng thw i se around sma ller d iameter p i pes and a l lowed to
coo l . Th e ne w shape ha d , as c l o se ly as was poss ib l e , a cons i s t en t
cross-se c t ional shape along the l eng th of the p i p e . The c r oss-se c t i o ns
o f the var i ous shapes produ ced are g i ven in F i g . 5 . 1 5 to 5 . 2 1 .

Hea t transfer along the l eng th of each ob j e c t was m i n i m i s ed in a


s i mi l ar mann e r to th at used for inf ini t e c y l inders .

The i r r e gular shap ed obj ec t s were f i l led w i th Ty lose or minced l e an


bee f fr om on e end to a mid-po i nt of the i r he ight . Thermocoup l e s were
pos i t ion ed at f i v e or s ix lo cat ions throu gho ut th e obj e c ts at th is
c entral leve l . Th i s was don e b y d r i l l i ng sma l l ho l es i n th e P VC p ipe
wall j u s t large enough for a s i ngle thermo couple lead. By inser t ing
the coppe r and const antan leads through sepa r a te ho l e s and reconn ec t i ng
and tens ion i ng th em w i th in the obj e c t as shown i n Fi g . 5 . 22 , the
ther mocoup l e j unc t ion was accurate l y pos i t ioned . The th ermo c oup l e
leads were held in place and the holes i n the p ipe wal ls were se aled
wi th a small amoun t of epoxy r es in . Pack i ng o f the ob j ec t w i th Ty l ose
was then completed and the poly styrene foam end c ap s were placed i n
pos i t io n . Th e loca t io n s of th e thermocouples for each shap e are
i nd i c at e d on Fi g . 5 . 1 5 to 5 . 21 . A f i lled t es t samp l e i s shown i n
F i g . 5 . 23 .
E xper imental Pr ocedur e and Dat a Co llect ion 99

ONE -DIMENSIONAL GRID

TWO - DIMENSIONAL GRID

S c ale A [cm]
0 1 2
S c ale B [cm]
1 2

F i g . 5 . 1 5 Cross - sec t io n and F ini te Elemen t Me thod Gr i ds For the


Two - D i me ns iona l Ir regular Shapes Numbers On e and Fi v e .
Sca le A appl i e s for Shape No . 1 . Sca l e B appl i e s for Shape
No . 5 . * - nodes , * * - nodes c orrespond ing to th ermocouples
pos i t ions , - e leme nt bounda r i e s , - - p l an es o f symme t ry .

MASSEY UNIVERS itY,


U B . A r' Y
Exper i mental Proc edure and Da ta Co l l ec t ion 1 00

Scale [cm]
4

Fig . 5.16 Cross-sec t io n and Fi n i te El e me n t M e th o d G r i d For the

Two- D i mens i o n a l I r r e gu l a r S hape Nu mbe r Two .

* - n od e s , * * - n od e s c o r r e s p o nd i n g t o t h e rmo c o up l e s

pos i t ions , - e l emen t boundar i e s .

Fig . 5 . 1 7 Cross-se c t ion and Fi n i te Eleme n t Me thod Gr i d For th e


Two-Di men s ional Irr egular Shape Number Thr ee .
* - node s , * * - nodes c orrespond ing t o thermo co uples
po s i t i ons , - e l e men t bounda r i e s .
E x per imental Pr oc edure and Dat a Co l l ec t i on 101

Scale [cm]
2

Fig. 5 . 1 8 C r o s s - s e c t i on and F i n i t e E l e me n t M e th od G r i d F o r t h e

T w o - D i me n s i o na l I r r eg u l a r Sh a p e N u mb e r Fo u r .

* - n od e s , ** - n o d e s c o r r e s p o nd i ng to t he r mo c o u p l e s

p os i t i o n s , - e l e me n t b ou n d a r i e s .

S c ale [cm]

F i g . 5 . 1 9 Cross-sect ion and F i n i te Elemen t M ethod Gr id For the


Two-D i mens ional Irregular Shap e Number S i x .
* - nodes , * * - nodes corre spond i ng t o thermocoupl e s
pos i t i ons , - e l ement boundar i e s .
Expe r i mental Pr ocedure and Da ta Co l l ec t i on 1 02

S c ale [cm]
-'--
'----- -L

F i g . 5 . 20 C r oss-se c t ion and F i n i t e El eme n t Method Gr id For th e


Two-Di me ns iona l I r re gular Shape Nu mb er Se ven .
* - nod es , ** - nodes c or respond ing to th ermoc o up les
po si t ions , - e l emen t boundar i e s .
S c ale [cm]
L___ _L __�'--- -�--�

Fig . 5 . 2 1 Cr oss-se c t i o n and Fi n it e El eme n t Me thod Gr i d For th e


Two- D imens i o nal Irr egu l ar Sh ape Nu mber Ei ght .
* - nodes , * * - nodes correspond ing t o th ermo co uples
pos i t ions , - e l emen t boundar i e s .
Exper imen t al Pr ocedur e and Dat a Co l le c t ion 1 03

c D E

F i g . 5 . 2 2 Sch ematic Di agram Show i ng the Me thod of The r mocoup l e


Ins ert ion and Pos i t ion ing W i th in the Mu l t i -Di mens i o na l
Irr egular Shapes . Not to sca l e .
A - object wa l l , B - epoxy res i n , C - insulated co ns tantan
lea d , D - the rmocoup l e j unc t i on , E - i nsul ?. t c d coppe r l ead .

Fi g . 5 . 23 The Sa mp le Osc i l la tor and Two- D i mens ional Ir r egular Shap e


Freez i ng and Thaw i ng Equ ip ment U se d in the L i qu id Immers i on
Tank .
Experi mental Procedure and Data Co l l ec t i on 1 04

The l i qu id imme r s i o n t ank was use d . Bec ause b o th fr eez ing and thawi ng
expe r i men t s were ca rr i ed ou t , th e tank cont a ined 29% calc ium ch l or i de
b r i ne . Ac c urate temperature control across th e tempe rature range - 4 0 °C
t o 5 0 °C was ach ieved. The test samp les wer e osc i l lated to m i n i mi s e
l oca l v ar i a t ions o f the sur fac e he a t trans f er coe ff ic i e nt .

5 .8 .2 D i mens i o nal Me asur emen t and Co ntrol

The r i g id nature of the PVC p i pes mea nt tha t no var i a t i o n in shape or


s iz e was me asured dur i ng ex p er i men t s . I f any densi t y change of the
Ty lose or m i nced lean beef occurred dur ing phase chang e , it resu lt e d i n
l e ng thwi se cha nge i n vo lume wh i c h was absorbed by compres s io n of the
poly styrene end cap s .

There was some variat ion in th e cross-se c t i onal shap e of each ob j e c t


a l o ng i ts l eng th . Cross-sec t iona l prof i l es fr om each end o f the obj ec t
wer e drawn and averaged to es t i ma te th e f i nal shap e and s i z e for e ach
obj ec t ( F i g . 5 . 1 5 to 5 . 2 1 ) as acc urate l y as pos s i b l e . V ar i at i ons f rom
the mean shap e were l e ss than ±1 . Omm in d i me ns ions and less than 1 . 0 %
i n per i me t er and cross-sec t ional a r ea .

The charac ter i s t i c th ick ness for th e i r regu lar shap es is def ined as
e i th er the minimum th i cknes s measur ed throu gh the the rmodynam i c centre
or double th e minimum d is tanc e from the the rmo d ynami c c en tre to th e
sur fac e . For al l th e shapes i t was found tha t on l y one th ermo dynamic
centre ex i s t e d , because the surface cond it ions were u n i form a n d th e
geome t r i e s u sed were re l a t i ve l y uni form in sh ape . Th e cha rac t er i st i c
th icknesses determined fr om each def in i t ion were equa l . The pos i t i o n
o f th e the rmod y nami c c en tre w a s found by ana ly s i s us i ng the f in i te
e lement me thod (Se c . 4 . 2 ) for typ ical phase c hange cond i t i o ns .

5 .8 .3 M ea suremen t and Contro l of Surface He at Transfer Coe f f ic i ents

For the i r r egul ar shapes s tud i e d there wer e no sat i sfactory me thods to
est ima t e th e sur fa c e heat transfer coeff ic ient d ir ec t ly from
exper i men ts for each shape . The two-d i mens iona l irregu l ar shapes were
r emou lded from PVC c y l inders for wh ich surfac e he at transfer
Experimental Pr oc edure and Dat a Co llection 1 05

coe f f ic i en t s c a n b e de t erm ined ( Sec . 5. 5 .3) . As l ong as the l i qu i d


f i l m heat transfer cond i t ions are equ ivalent at all par ts o f the
sur fac e and th e P VC wal l th i ckness i s cons t a nt ov er th e who l e surfa ce ,
then th e surface heat tr ansfer coeff i c ient wou ld be expected to be
approx i ma t e l y th e same as that for th e equ iva l en t P VC cy l ind er . U se o f
the l i q u i d i mmer s io n sy stem plus osc illat ion of th e tes t sa mples
ensure d that th e l i qu i d f i l m heat trans f er coeff ic i e nt wa s constant to
all surfaces of all th e objects . Th e PVC wall th ick ness was th e sa me
for th e remoulded sh apes as for th e orig ina l cyl inder s and d i d no t vary
s i g n i f icantly at d i fferent pos i t ions on the surface of th e i r r egular
shapes . Smal l random var i a t ions in wa l l th i ckness d id occur d ue to
chang es in th e degree of curv ature of the PVC and heat shr i nk age . The
ins i d e pe r i meter s o f th e irregular shapes we re measured to b e w i th i n 1 %
of the or ig inal c y l i nder wall perime ter for the 0. 15 m cy l i nder s and
wi th i n 2% for the 0.1 m cyl inder s , for all th e irregu l ar shapes .
Therefore var i a t i o n i n the PVC wa ll th ick ness due to st re tch ing dur ing
the mould i ng proc es s ha d n egl i g i b l e ef fec t on h eat trans f er .

Separat e heat i ng and coo l i ng experiments w i th an unmou lded P VC c y l inder


we re u sed to es t ima te h for th e oth er ( irregula r ) shapes i n a s i m i l ar
manner to that use d in infinite cyl i nder thawi ng ex per i ments
( Sec . 5 .5 .3 ) . Th e curved sur fac es o f both the irr egul ar shap es a nd the
cyl i nder gave the same prob lems with voids and imperfect th erma l
contac t a t th e sur fac e , ex p er i en ced for inf i n i te cy l inde r and s phere
thaw i ng . It was assume d that th e effec t o f th ese fac tor s were th e sa me
for the irregular shapes as for the cyl ind er shapes so tha t the
est imated surface heat transfer coef f ic ien ts were s t i l l r ep resenta t i v e .
B e cau se Ty l ose pa ck i ng cond i t ions and the occur r ance and d i s tr i bu t i on
of the surface voids we re s i mi lar i n both cases , there i s no r eason to
expec t th i s assump t i on t o be i nva l i d .

To alter the surface heat transfer coeff ic ient two lay er s of sheet
rubber were g lued onto the sur fac e of a ll the irregul ar l y shap ed
obj e cts der i v ed fr om the 0. 1 m PVC p ipes . The chang e in exposed
sur fac e a r ea as mor e packa g i ng mate r i a l is add ed is sl i gh t l y d i f f er e nt
for a c y l inder and an i rregular shap e , so very small d i f ferenc es in h
values are poss i b l e , but these were w i th in the toleran ce o f the
measureme n t system. The e ffec t of the rubber and the abov e prob lems
Expe r i mental Procedure and Data Col l ect ion 106

w i th i mperfec t c on tac t are d isc ussed f u l ly i n Sec . 5 . 5 . 3 .

Th e heat penetra t io n me thod , bas ed on cyl ind er c en t r e t emperature


prof i les , showe d no stat i s t i c al ly s i gn i f i c an t d if ferenc es b e twe en
r ep l i cates wi th in the 26 runs used to es t i ma te the thr ee d i f f erent
values of h ( for the 0 . 1 5 m PVC c y l i nder , the 0 . 1 m PVC cyl inder and
the 0 . 1 m P VC cyl inder p lu s two l ayer s of rubb er ) at the 95% leve l of
con f i d enc e . The unc erta i n ty in the mean values had 95% conf idence
bounds of ±1 2 . 7 % . ±7 . 1 % and ±2.8% for the abov e three va lues
respec t i ve l y . Th e va lues of h were d i f ferent fr om those for the 0 .1 m
and 0 . 1 5 m PVC cyl inders u sed in thaw i ng ( Sec . 5 . 5 ) because p ipes of
d i fferent th i ckn es ses were u sed .

The sur face heat trans fer coe f f ic ient for the minc ed l e an bee f
expe r i men t s was assu med to be the same as that de termine d for the
Tylose exp e r i me n ts w i th th e same shap e . As d iscu ssed in Sec . 4 . 8 . 3 for
rectangula r b r i cks , th i s poten t ia l ly l ead s to in creased ex pe r imenta l
uncer ta inty for the minced lean beef experiments b ecause th is
assu mp t ion abou t h may no t b e va l i d .

5 .8 .4 Analy s i s o f Heat Tr ansfer i n Two- D i mens ional Irregular Shapes

Experiments for i r r e gular shapes are subj ec t to the same e rror to thos e
for regu l ar shapes . Vo ids in the Ty lo se and inhomogenei t y cau sed by
thermoc oup l e wires had s i mi lar e ffects to those found in the i n f in i t e
cyl inder expe r i men t s , and wer e s i mi lar ly ins ignif ic ant .

Error i n the rmocou ple placemen t was not a major sourc e of unc er tainty
due to th e way the le ads were po si t ion ed and he l d in p l ace .
D i smantl i ng of th e obj e c t s showed that the thermo couple j unc t i o ns
rema ined wi th i n a 2 mm rad ius of th e recorded pos i t io n ov er a numb er of
runs . However , the me thod used to i nse rt th e thermocouples d id l e ad to
error s du e to heat conduc t ion a lo ng the w i res as they were not
i nt roduc ed along isoth ermal p aths . Ty pically , four th ermo co uples were
used to measure in terna l t emperatures for each obj ec t . On l y o ne of
these was located in close prox imity to th e thermodynamic c en tr e .
Conduc t ion along e ach thermocoupl e wou l d b e l es s than th e 0 . 8% o f the
total heat flow calc ulated for the centre thermocouple i n the sma llest
Exper imental Pr ocedure and Da ta Co llect ion 1 07

sphere u se d for thaw i ng ( Sec . 5 . 6 . 4 ) . The vo lume and sur fac e ar ea for
these two -d ime ns ional i r re gular shap es wer e l a rger that tho s e for the
sma l l es t spher e . Though up to four thermocoup l e w i res conducted heat
towar ds the cen tre , because all but one of th ese were pos i t ioned well
away from th e thermod ynamic c entre and t he ob j ec t vo lu me and sur f ace
ar eas were larg e , the e ffec t of heat co nduc t ion a long th e the rmo couple
w i res was cons idered ins i gn i f ican t . On ly 0 . 0 2 % to 0 . 0 3% of th e sur fac e
ar ea w i th in the c en tra l 1 00 mm l eng th of the p ipes was a ffected by th e
introduc t ion of th e th ermo c ouple w i r e throug ht the wa l l s . Th i s is
insign i f icantly sma l l .

Hea t transfer a long the l eng th of the ir regularly shap ed obj ec ts will
be of th e same ord er of magni tud e as that for th e equ i va l ent
cyl indr i c a l shapes due to th e insu lat ing cap s , th e h igh rat io of leng th
to th i ck n ess u sed a nd th e cen t ral pos i t io n used for temperature
mea surement . In Sec . 5 . 5 . 4 th is was shown to b e ins ign if ican t .

Th e prob l em wi th i mpe rfec t contac t of the Ty lose w i th th e P VC wa l l ,


noted for in f i n i t e cyl inder experiments , occurre d . For i n f in i te
cy l i nders it was ana ly sed in the same manne r by us i ng c entre
t emperature prof i les only to determine an ave rage h for each obj e c t .
The e f f ec t of sur fac e inhomogenei ty , due to the imperfec t contac t , may
have b een larger for i r regular shapes i f the presenc e of voids was no t
even l y d i str ibu ted ov er th e sur fac e . There was no ev i den ce o f th i s so
the approach use d for i n f i n i t e c y l inders when cons ider ing accuracy of
sur fac e temperature p rof iles was adop ted .

For i r r e gular shap es the pos i t ion of the the rmodynamic c en tre is not
k nown , and in f ac t it may mov e dur i ng a ny fr eez ing or tha w i ng run
( Fl eming 1 9 7 0 ) . For th is r e ason , th e pos i t ion of one the rmo couple was
located as c lo se as poss i b l e to th e pred i c t e d thermody nami c c entre of
the obj e c t s , as de t ermined by a f in i te e lement analy s i s ( Se c . 4 . 2 ) for
typ ic a l ex per i men ta l cond i t ions . Th i s may no t be the ex ac t
thermo dy namic c en tre so when the temperature p r of i les were analy se d ,
the b r e akpo in t ana lys i s d i scuss e d i n Sec . 5 . 4 . 4 was u se d t o increa se
the prec is io n o f th e exper ime n ta l ly determined thawing t i me where
n ec essary . Th i s ana lys i s i s as app l i cabl e to irregu la r shape s as it i s
to regular shapes . When c o ns ider i ng loc a t ions o f oth er the rmo co uples
Exper i ment a l Procedure and Da ta Col l ec t io n 1 08

i t was no t i mpor t an t where th ey were pos i t ioned as long as the pos i t ion
was acc urat e l y k nown . A r ange of a l terna t i v e posi t ions we r e u sed to
prov ide a numb er of ty p ical t emperature profi les f r om th rou ghout each
ob j ec t .

For the two-d ime ns ional i r r e gular shape exper iments an iden ti c al
p rocedure to that u sed for thaw i ng o f regu l ar in f ini te cyl inder s was
adopted . Al thou gh the unc erta inty in measured sur fac e heat t ransfer
coe ff ic ients was lowe r than for th e inf in i te cyl inder th aw i ng
exper ime nt s , the inab i l i ty to control h to th e same value for all the
sh ap es introduced ex t r a unc er ta inty . The re i s a lso an add it iona l er ror
in measur ing and control l i ng the geometry so the ov e ra l l ex per i mental
error was es t i mated to b e of sl ight l y gre a t er magn i tude than tha t for
i n f i n i t e cyl inder thaw ing .

5 .9 H EAT TRANSFER I N TH REE- DIMEN SIONAL I R R EGU LAR SHAPES

5.9 .1 The Equ ipment

On ly thr e e d i fferent thr ee - d i mens ional obj e c t s were inv e s t igated due to
l i mi t a t ions in t i me a nd r es ourc e s . Each was rela t i v e l y un i form in
shap e , though i r regular compar ed w i th slab s , inf in i te c y l i nder s ,
spheres , inf ini t e rods or r ect angular br i ck s .

The f ir s t shap e was th e fru stum o f a sq uar e pyramid shown in - Fi g s . 5 . 2 4


and 5 . 28 . I t was const ruc t ed from po l yp r opy l e ne plast i c she et in
exac t ly the same manner as that used for rec t angular b r ick shap es
( Sec . 5 . 7 . 1 ) . The o ther two irregula r thr ee - d i mensiona l shapes w ere an
ov o i d ( eg g ) shape and an f i sh shap e smoothed to ob tain a comp letely
convex sur fac e . To prov id e in forma t ion ab ou t sur fa ce he at t r ansf er
cond i t ions , two s i mi l a rly s i z ed r egular spheres were use d . Casts of
each shape wer e made . For th e eg g shap e , a t wo-d imens i ona l
cross - sec t i onal p r of i le was d rawn and a b lo ck o f wood turned i n a l a th e
t o th i s shap e . The ful l wooden egg was se t i n p l ast er o f par i s so that
t wo iden t i c a l ha lves c ou ld be mou lded. The plaster prov ided a flat
surfac e t o form a f l ange so the two ha lves c ou l d b e j o in e d t ogether .
An albacore tuna was f r oz en and th en c u t i n half leng th w i se down the
b ac kbone . A n ega t i v e p l as t er o f par i s cast was then made o f one hal f .
Experimental Pr ocedur e and Dat a Co llec t i on 1 09

From th i s posi t i ve f i sh shaped cas t s were r econs t ructed . Th e spher i ca l


plaster of par i s c as t was ma de in the hollow metal h emi spheres u se d to
make the 0 . 1 28 m d iame t e r sphere for thaw i ng of r egul ar shapes . The
me tal sphere was not su itab le for mou ld ing d ir ectly .

The c as t s were use d to vacuum he a t mou l d t wo ha lves for each ob j ec t


from C A B ethyl a c etate p last i c sh eet . The f i sh , eg g and two sphere
mou l d s const ruct ed were a l l of compa rab l e s iz e and curv a ture so that
the vac uum mould ing of the plast i c gave s i mi lar th ick nesses of p last i c
over e ach of the moulds . Th e cas ts ha d no conc ave sect ions so that
b r i dg ing of the pla s t i c between r i dges on th e cas t was no t a problem
and the plas t i c mou l d c lo se ly resemb l ed and f i t ted the or ig ina l cast .

Each pla s t i c half mould was f i l le d w i th Ty lose . Thermocouples were


introduced to v ar i ous pre- de termined po in ts i n a s i mi l ar manner to that
used for two-d imens ional irregular shapes . Holes were d r i l led in the
plas t ic for the ind i v i dua l copper or const an tan w ir e l e ads . The w i res
wer e j o i n ed and tens io ned in the approp r iate place in the shap es and
th e ho l es sea l e d by epo xy r es i n . Six the rmocouples ; two on the
surface and four located i nt ernally , wer e use d for each shape , ex cep t
for th e two sph eres wh ich each had thr ee sur fac e thermoc oup les a nd one
at th e geome t r i c cen tre . One the rmo coup le was always plac ed as c lose
as po ssible to the thermodynamic c entre est i ma ted by f i ni te e l ement
ana ly s is . The f i l led halves of th e plas t i c objects were j o ined
toge th er and th e plas t ic fl anges he l d toge th er w i th a so lven t g l ue and
sc rews . In th is way th e obj e c t s were totally sealed aga ins t l i quid
i ngres s .

The thr ee shapes are shown i n Fi g . 5 . 25 to 5 . 28 . Al l the shap es had a


bo l t att ach ed throu gh th e flange to allow conn ec t ion to th e samp l e
osc i l lator . The l i q u id imme r sion tank was u se d .

5 .9 . 2 D i mensiona l Control and Measurement

The measuremen t s o f the moulded p l as t i c ob j ec t s wer e t ak en d irec t ly


from th e wood or plas ter cas ts . The wooden egg c los ely fo l lowed th e
mathema t ical re lat ionsh i p for a n ovo i d b ecau se i t was tur n e d to fit a
spec i f ic p rof i le . S i mi larly , th e mou lds for the spheres and the
Exp er i mental Procedur e and Dat a Col l ec t ion 110

frustum of the square pyramid were uni form or flat s ided and cou ld be
accurate ly measur ed . Th e f i sh shape was natura l l y der i v ed and
irregular i n a l l r espects excep t for a leng thw i se plane of symme try
betwe en the two ha lves . To acc urate ly mode l i t , th e plaster ca st was
cu t and the c ross-s ec t i ons at th ir t een evenly d i s tr i b uted po i n ts a long
th e l eng th of th e ob j e c t mappe d . The sh ape of the f i sh sur f a ce was
then i n t erpolated between these c ros s-se c t i o ns . I t was found that each
of th e cross - sec t ions was c lo s l y appr ox i mated if a fur ther v er t i c a l
ax is of sy mme try was assume d . Hence only on e qua rter of the f i sh wa s
mode l le d . U s i ng these methods and be cau se the plas t i c mou l d c l ose l y
resemb le d t h e o r i g inal casts the error in de termin ing the sur face shap e
and c r i t i c al d imens ions was cons idered to b e ±2 mm a t any point for
both the f i sh and th e egg , ± 1 . 0 mm for the spheres and ±0 . 5 mm for the
py rami d .

The charac ter i s t i c th ick nes s was de termined from the grids shown in
Fig . 5 . 2 4 to 5 . 27 onc e th e the rmodynam i c c entre was p i npo in t ed by
f i n i t e e lement analy si s for typ ical phase change cond i ti ons . The
measuremen ts taken of f th e cas t s wer e conf irme d by some measurements on
the ac tual test obj e c t s . There was n o obse rv ab l e d i fferenc e in s iz e or
shape as the obj ec ts c hanged from th e fr oz en to th e un fr ozen s t a tes .

5 . 9 .3 Co n trol and Measurement of Sur fac e Heat Trans fer Co ef f i c ient

The polyp ropylene pyrami d was i den t i c a l in all respects to the


polypr opy l en e rect angu l ar br i ck sh apes cons t ruc ted prev iously for water
imme rs ion thawi ng ( Sec . 5 . 7 ) a nd for b r i n e immer sion freez ing ( Cl e land
1 977 ) , excep t for th e irregula r i t ies in sh ap e . Th e sur fac e heat
transfer coeff ic ient ha d been found to be equa l in b o th th e rectangular
b r i ck freez i ng e xper i men ts us i ng b r in e ( Cl e l and 1 977 ) and the thaw i ng
exper i ments us ing wate r . The sa me h v a lue was assume d t o hold for the
po lyp ropy l e ne pyram i d and no fur the r experi ments re l at i ng to i t we re
under tak en . The er ror i n the value of h for the poly propylene pyramid
sh ou l d be of th e same order as tha t for tha w i ng of the po l yp ropy l ene
rectangular b rick s ( 1 . 5% ) . The samp l e was osc i l lated in a s i mi la r
manner t o tha t used for rectangular b r i ck s .
E xper imental Pr ocedur e and Dat a Co l le c t i o n 111

For th e p last i c mou ld shapes, h cannot be measured d irec t l y w i th ou t


substan t i a l d i f f ic ulty . Henc e the analagou s regula r spher i c al shap es
construc t e d fr om the same plas t i c mater ia l , and th e me thods of
Se c . 5 . 6 . 3 , were use d to measure h. Twe lve separate heat ing and
coo l i ng runs w i thout phase change , b oth ab ove and be l ow th e pha se
change t emp erature rang e , for the two sphere shap es were undertak en .
Non e o f the d i ffer enc es bet ween runs , be tween v a lu es for the two
spheres or th e corre lat ion of h w i th temperature were s tat i s t i c al ly
s i gn i f i cant at th e 95 % conf i denc e leve l . The err or in the average
value determ i n ed by the va riab l i lty of r e p l i c ates was ±21 % . It was
assumed that th e h v a lu e de termined in th i s way accura te l y r ep re sented
h for the egg and f ish shapes as wa l l th ick nesses were essen t i a l ly the
same for all t es t ob j ect s , and th e degre es of curv a ture were
ap p rox ima t e ly equiva lent . Some plas t i c th ickness variation was no te d ,
e ven between th e two spheres moulded from i de nt ic a l cas ts , d ue to
fa c tor s tha t were no t controllab l e . Dur i ng mould ing the plas t i c tended
to sp re ad ou t th inly over some r eg ions and accumu l ate at ot her parts .
The effec t was rela t i ve ly randomly d i s t r i buted excep t nea r the flanges
for each ha l f wh ere the plas t i c was a lway s th i ck er . Henc e , the re was
in p rac t i c e some v io lat ion of th e assump tion th at ex te rna l heat
t r ansfer was un i form on a l l sur faces of th e test ob j ect s . No pract i ca l
way t o avoid th is p rob lem was found , so i t was dec ided t o carry ou t th e
exper iments wi th the moulds as th ey wer e , and acc ep t the consequ e nt i a l
l o s s of ac curacy in calculat io ns for wh ich a cons tant h value was
assume d .

5 .9 . 4 Ana ly s is of Hea t Tr ansfer for Three-D i mensional Ir regu la r Shapes

Sou rces of error for th e three-d i mens ional i r r egular shap es were
simi l ar to th ose for o th er sh apes . Pr ob lems wi th vo ids , inhomogene i t y
d u e to thermo co uple w i r es , heat conduct ion along thermo coup l e w i r es and
error s in p lac emen t of thermocoup l es wer e s i mi l ar to those for
two - d imens ional irregular shapes and were cons equen t ly c ons idered
n eg l i g ib l e . Becau se a the rmocoupl e c ou l d no t neces sari l y be lo cated
exac tly a t th e thermo dynamic c en tre ( but usu a l ly c lose by) , th e
b r eakpo in t ana ly s i s was u sed to accurate l y de t e rm in e ex p er i menta l
freez ing a nd thaw i ng t i mes .
E xpe r i me n tal Pr oc edure and Dat a Co llect ion 112

The three-d i mensiona l sh apes mean t tha t end he at transfer e f fec ts were
no t a prob lem. On ly the pyramid shap e was a ffected by the prese nc e o f
a bo l t inser t e d through the wa l l . As was th e ca se for rectangu l ar
brick thaw i ng , the p resenc e o f th is bolt was found to be ins ign i f ican t .
D i srup t ions due to th e modu l ar nature o f the p l as t i c mou lds a f f ec t ed
hea t trans f e r . The flange reta ined on each hal f of th e plas t i c mou lds
a l lowe d a s trong j o i n to be ma de a nd i t a lso a l lowe d th e att achme nt to
the sample osc i l la tor w i thou t fur ther d isrup tion to the obj e c t sur fac e .
The f la nge for the plas t i c mou l ded shapes ran in a l e ng thw i se
c ircumferenc e around each shap e . It therefore af fec ted part o f th e
sur fa ce area but the reg ions on th e sur face c l oses t to the
thermod ynam ic cen tre were only s l i g htly a f fec ted . The e f fec t of the
flange on the sur fac e h eat transfe r was cons idered unavoi dabl e . As has
bee n ind ica t e d , a lower d e gree of ac c uracy in th ese expe r i me n ts was
acc ep ted .

Imperfect co ntac t was a s i mi lar p rob lem for th e pla s t ic mou lded shapes
to that wh ich occ ur r ed for inf ini t e cyl inder s , spheres a nd the
two-d ime ns iona l ly ir regular shapes . As b e for e , it was cons i dered that
the e f fect was equa l ly fr equ en t for a l l sh apes a nd rand om l y d i s tr i bu ted
ov er th e surface so that th e ef fec t on surface or ov erall h eat transfer
was not sign i f i cant and cou l d b e acc urate l y ac coun t e d for by an a v erage
sur fac e heat transfer coeff ic ien t . Fi n i te e lement pred i c t i o ns for the
p las t i c shapes ( Sec . 8 . 4 ) tend to show tha t surface he at trans f er
cond i t io ns were no t un i form and that alter nat i v e me ans to con trol and
measur e h wou l d be need ed i f ex peri men ts wi th l e ss unc erta in t y were
con s i d ered esse n t i a l . In sp i t e of these problems , the shape o f the
temperature p rof i l e for a th ermocoup l e s i tuated near the thermo dyna mic
centre ( pred ic ted by f i n i te eleme n t analy s i s assuming un i form surface
h ea t t r ansfe r cond i t ions ) , sugges ted tha t th e non-un i formi t i es at the
sur fac e had not mov e d the th ermo dynamic centre much , and that thaw ing
t i me s cou l d be de term i ned from th i s th ermocoupl e by the br eakpo in t
analy s i s .

For three-d ime ns ional i r regu lar shap es the unc er taJnty assoc iated w i th
contro l l i ng the sur fac e hea t transfer cond i t ions by u se o f the plast i c
mou ld ing could not b e assessed . However the error i n me asur i ng h was
large and ther e are furth er s ign i f icant error s in measur i ng and
E xp e r imen tal Pr ocedure and Dat a Co l lec t io n 113

model l i ng th e shapes o f each ob j e c t accurat e l y . Other s ou r ces of


uncer t a i n ty a r e of the same order o f magn i tude t o those in th e o th er
exper i men ts . Co nsequ e nt l y , the ov era l l ex p er i menta l err or f or the
thr ee -d i mens ionall y i r r e gular shapes mou lded fr om plas t i c was larger
than for o th er shapes . Th i s mu st b e cons i dered wh en compa r i ng freez i ng
and thaw i ng t i me pre d i c t i o n me thod s w i th the e x per imental data. The
excep t io n was th e polypr opy l ene pyram i d sh ape for wh i ch the ov era l l
exper i mental error was o f th e same order o f magn i tude a s occurr ed for
the thaw i ng of rec tangular b r i c k s as the ex per i menta l t echniques use d
were a l mo s t i dent i ca l .
Exper imental Pr ocedure and Da t a Co l lec t io n 114

Fi g . 5 . 2 4 The Pyramid Three-Di mens ional Irregu l ar Sh ape F in i t e El ement


Me thod Gr i d .
Only a q uadrant o f th e shap e i s mod e l l e d . T h e x , z and y , z
fac e s a r e p lanes of symme t ry . On ly the boundar y sur face
grids are shown . The x , y face i s also a b oundary surface.
Sca le - in the x d irec t ion 1 d iv is ion = 0 . 005 m , in the y
d ir ec t i o n 1 d i v is ion = 0 . 008 m and i n th e z d ir ec t ion 1
d iv i s ion = 0 . 0 1 6 m .
Exper imental Pr ocedure and Dat a Co l lec t io n 1 15

ONE - DIMENSIONAL GRID


, ___. � � � � � � ---- � � •2
a

TWO -DIMENSIONAL GRID

S c ale [ c m ]
1.__
--L.
_ _J_
_ t_____j

0 1 2

F i g . 5 . 2 5 The Sphere Three- D i mens ional Irregular Shape F in i te Ele me nt


Me thod Gr ids .
( a ) and ( b ) the one- and two-d imens ional g r i d s
* - nodes , * * - nodes correspond i ng t o the rmocoupl e
pos i tions , - e lement bou ndaries , � - p lan es of symme t ry ,
A - ax i s o f rotat iona l sy mme t ry .
( c ) th e thr ee-d imens ional g r i d ( p . 1 1 6 ) .
E x pe r i mental Proc edure and Dat a Col l ec t i on 1 16

Fig . 5 . 25 The Sphere Three-d imens ional Irr eg ular Shap e Fi n i t e El eme n t
Me thod Gr ids .
( a ) and ( b ) the one- and two - d ime ns ional g r i d s ( p . 1 1 5 )
( c ) the thr ee - d i mens iona l gr i d . On l y a n oct ant i s mode l l ed .
Th e x , z and y , z faces are planes o f symmetry . Scale for the
thr ee-d imens iona l gr i d is 1 d iv is ion - 0 . 00635 m .
( 1 ) only the boundary surfaces are shown.
( 1 1 ) an exploded v i ew o f the ful l spher e gr i d ; each o f the
thr e e she l ls o f e lements are shown.
Exper imental Proc edure and Da ta Co l lect ion 1 17

TWO - DIMENSIONAL GRID

2 4

THREE- DIMENSIONAL GRID

F i g . 5 . 2 6 The Egg Thr ee-D i mens iona l Irregular Shape F in i t e El ement


Me thod Gr ids .
( a ) th e two-d imens iona l g r i d . * - nodes , ** - nodes
corre spond ing to thermocoup l e pos i t ions , - e l eme nt
boundar ies , A - ax i s of ro tat ional symmetry .
( b ) the three-d i mens iona l gr i d . On l y a quad rant i s mode l l ed .
The x , y and x , z faces are planes o f symme try . Only the
b oundary sur fac e gr i d is shown . The fu l l g r i d is an
e longated , doub l e vers i o n o f the sphere thr e e- d imens ional
g r id ( F i g . 5 . 25 c ) . Sca l es : in th e x d irect ion 1 d i v is ion
0 . 0 2 55 m, in the y and z d ir ec t ions d i v i s ion = 0 . 0085 m.
E x per i mental Procedure and Data Co l l ec t i on 1 18
z

Fi g . 5 . 2 7 The Fi sh Three-D i mens ional Irr egular Shape Fi n i te El eme n t


Method Gr i d . On l y a qu adra nt i s mode l led . The x, y a nd x , z
faces are planes of symme try . On ly the bounda ry sur fac e gr i d
i s shown . The fu l l gr i d i s a d i s t or ted versio n o f th e egg
three-d i mens ional grid ( F i g . 5 . 26 b ) . Scales : in th e x
d irect ion 1 d iv is ion 0 . 0 60 m , in the y d ir ec t i on 1 d iv i s ion
= 0 . 007 m and in th e z d ir ec t io n 1 d iv is ion = 0 . 006 m.
( a ) v iewe d fr om th e t ai l , ( b ) v iewe d from the he ad .
Exper ime ntal Pr ocedure and Dat a Co l lectio n 1 19

b
F i g . 5 . 28 The Sa mp le Osc i l lator and Three- Di mens iona l Irregu lar Sh ap es
Use d i n the Liquid Imme r s i o n Tank .
( a ) pyrami d
( b ) sphere
( c ) egg ( p. 1 20 )
( d ) f i sh ( p . 1 20 ) .
Exper imental Pr ocedure and Dat a Co l lec tion 1 20

• •

d •

F i g . 5 . 28 The Sample Osc i l lator and Three-Di mens ional Irr egular Sh apes
Used in the Liquid Immers ion Tank .
( a ) pyramid ( p . 1 19 )

( b ) sphere ( p . 1 19)
( c ) egg
( d ) f i sh
121

6 EXP ER IMENTAL DESIGN A ND R ESULTS

6.1 INTRODUCT ION

To inves t i gate the accuracy of methods for pre d i c t i ng freez i ng and


thawi ng t imes for food , exper iments must be conduc t ed for a wide range
of cond i t i ons cover i ng those that occur in prac t i ca l freez i ng and
thaw i ng si tuat ions . In Chap . 3 f i ve obj ec t i ves for the p r es e nt wor k
wer e defi ned . The f i rst two of these re lated to data coll ec t i on so
that pred i c t ion methods could be ass essed :
( 1 ) to co l l ec t thaw i ng data for the bas i c s l ab , infi n i t e cyl i nd er ,
s phere and rec tangular b r i ck shap es
( 2 ) to collect fre e z i ng and thaw ing data for other ( i rregular ) shapes .
I t was dec i ded to cons i der each shape separa tely and to l i mi t the range
of freez ing and thaw ing cond i t ions to si tua t ions descr i bed in Tabl e
3.2. Essent i al l y , th i s meant that the study was l i mi ted t o homogeneous
mater i als and constant cond it ions . Even aft er these restr i ct i ons on
the range of cond i t ions are appl i e d , seven important factors affec t i ng
freez i ng or thaw i ng t ime can b e iden t i f ied ( Tabl e 3 . 3 ) .

Pham ( 1 983 ) used d imen s i onal analys i s to show that the six factors
other than shap e can be related by seven d imens ion l ess group s t o the
freez i ng or thaw ing t i me . These are the Four i er numb er , the B i ot
numb er , the Stefan number , the Plank number and three other
d imens ion l ess numbers . The Four i er numb er :

(6 . 1 )

t akes account of the obj e c t s i ze , the phase change t ime and the thermal
proper t i es of the mater ial in the state i t wi l l be i n after the phase
change p rocess has occurred . The Biot number :
hD hD
B i - - for fr eez i ng , B i for thawi ng
- -
(6 . 2 )
ks kl
i s the rat i o of i nt ernal t o external resi stances to heat transfer . The
Stefan number :

(6 . 3 )
Exper i menta l Des ign and Resul t s 1 22

accounts for the amb ient med i um temperature and re lates the heat
transfer after the phase change process to that due to latent heat
release or absor p t i o n i t se l f . The Plank number ( Cl e land & Earl e
1 976b ) :
C l ( T i n- T i f ) Cs ( T i f- Ti n )
Pk = for freez i ng , Pk for tha w i ng (6.4)
=

�H �
t ak es account o f the in i t ial superheat i ng or subcoo l i ng and relates i t
to the lat ent heat . F i na l ly , three other numb ers wh i ch take account o f
the f i nal the thermodynamic centre , T f i n • the i n i t ia l
temperature at
freez i ng temperature and o ther therma l propert i es of the food are :
C s ( T i f - Tf i n ) C l ( Tf i n - T i f )
N1 = for freez ing , N 1 = for thawi ng ( 6 .5 )
� �
cl kl
N2 = -- and N 3 = -- ( 6 . 6 ) and ( 6 . 7 )
cs ks
I n s i tua t ions where i t i s assumed that all l at ent heat i s released at a
un i que phase change temperature the entha l py change i n the phase change
temperature reg ion , �H . is defined as the latent heat of free z i ng or
thaw i ng , whe reas the entha lpy change between 0 °C and - 1 0 °C is used i n
all other cases . Th is def i n i t ion ensures that � inc ludes the maj or
phase change effect yet i s s imi l ar l y de f i ned and equa l l y re l ev an t for
both fr e e z i ng and thaw i ng analyses . For freez ing , the very s i m i l ar
defi n i t i o n of � as the entha lpy change from T i f t o - 1 0 °C has been used
( Cl e land 1 9 7 7 ) . Th i s def i n i t ion l eads to � values only 1 . 0 % d i f f erent
from � from 0 °C to - 1 0 °C and is therefore treated as be ing equ ivalent
i n th i s s t udy . Th i s def ini t ion is a lso commensura te wi th an
a l terna t i ve used by Hung & Thompson ( 1 9 83 ) where � was tak en as the
entha l py change from T in to T fin ·

Both Hayak awa et a l ( 1 9 83 a ) and Succar & Hayakawa ( 1 9 84 ) per forme d


numer i ca l l y bas ed screen i ng exper iments that showed tha t for s l ab and
i n f i n i t e c y l i nder freez i ng over a range of typ i ca l l y encount ered
cond i t ions only f i ve parame ters , descr ibed in t erms of Bi , St e and Pk
abov e , had a stat i s t i c a l l y s i gn i f icant e ffect on the freez ing t i me
( Fo ) . Nei ther the thermodynami c centre t emperat ur e nor the rat i o o f
t h e unfroz en to the frozen thermal propert i es (N1, N2, N3) were
s i gn i f i cant . Simi lar r esul ts would b e expected for thawi ng . Cleland &
Earl e ( 1 9 8 4 b ) show that the effect o n freez i ng t ime o f d if f er ent f i na l
thermodynami c centre t emperatures can be accurate l y related to the
Exper imental Des i gn and Results 1 23

Stefan number in a s i mple manner . For these reasons , and those


out l i ned in Sec . 2 . 2 . 3 only freez i ng to a s i ngle f i nal thermodynamic
centre t emperature of - 1 0 °C and thawi ng to 0 °C were i nv es t i ga t e d
exper imenta l l y .

Mos t foods , frozen o r thawed induc tr i a lly ar e predominantly aqueous in


compos i t ion and consequently the in i t ia l freez i ng temperature , T i f ' and
the r a t i o s of the thermal proper t i es , N 2 and N 3 , d i ffer on ly s l i ghtly
for d i f f erent foods ( Tabl e 6 . 1 ) . For th is r eason on ly one t y p i cal
phase change mater i a l ( Ty lose ) was used and the changes in N2 and N3

were not cons i dered of maj or prac t i cal import anc e . Some runs w i th a
food mat e r i a l , minced l ean beef , were per formed to see if any
not i c eab l e effec t due to N2 or N3 could be observed . A study
encompass ing a range of mater ials ( C lel and & Ear l e 1 98 4 a ) showed l i t t l e
dev iat ion b etween pre d i c t ions for the d if ferent food mater i a ls used ,
con f i rm i ng th i s reason i ng .

The f u l l f unc t i onal dependence of the phase change t i me to a g i ven


end po i n t temperature , on the cond i t ions cons i dered in th i s work and the
therma l p roper t i es i s therefore g i ven by :
Fo = f ( Bi , Ste , P k , geometry ) (6 .8 )
The var i ab l es that were man i pu lated in order to inves t i gate th i s
relat ionsh i p were the s i z e , D , the sur fac e heat transfer coeff ic ient ,
h , the amb i ent med i um temperature , T a , the i n i t i a l t emperatur e , T in and
the obj e c t shap e . To isolate the e ff ec t of geometry the other four
var i ab l es wer e var i ed i ndependently for a var i e t y of d i f f eren t shapes .
Tab l e 6 . 1 shows t yp ical values of these var i ab l es for common free z i ng
and thawing processes . An exper i mental des i gn was s ought in wh i ch
these var i ab l es were var i ed over as w i de a range as poss i bl e w i t h in the
the l i mi t a t ions of the exper i mental equ i pmen t .

Orthogonal exper imental des i gns i n t erms of Bi , Ste and Pk wer e not
f eas i b l e b ecause of phy s ical l imi t at ions in contro l l i ng the f our
var iables , espec ially h and D , to p re-selected values . The obj e c t s i ze
and the surface heat transfer coeff ic i en t s wer e , in mo s t cases ,
determi ned more by con s i derat i ons such as the ava i l ab il i ty of the
appropr i at e types of conta i ner and the i r wal l mater ial , than by cho i ce
to give a spec i f i ed Biot numb er . var i at ion in Bi was there fore
E xp er i men ta l Des ign and R esults 1 24

d i f f i cu l t t o ach ieve and ev enly spaced l ev e ls requ ired for an


or thogonal des ign were not prac t i cab l e . Consequently i t was dec ided to
vary the Biot number as widely as poss i b l e by using d if ferent
comb i nat ions of conta iner s ize and wal l mat er ia l s . The add i t ion of
i nsulat i ng rubber lay er s for some shapes gave some control and
var i ab i l i ty of h .

A t each level o f Biot numb er Ta and T in ' and hence Ste and P k , wer e
a l so var ied to cover as w i de a range and comb inat ions of cond i t ions as
poss i b l e . The bas i s for the experimenta l des i gn in these cases wer e
e i ther part or ful lfactor ial des igns wi th two or three l evels of T a
and T i n ' I t was dec ided not to at temp t t o control these two var i ab l es
exac t ly to the pre - se l ec ted l ev e ls ; p rov ided a va lue c lose to the
pre- se l ec ted level was ob tained, i t was cons idered sa t i sfac tory .

6.2 THAW ING OF SL ABS

For exper iments u s i ng Tylose i t was chosen to var y h and D to g i ve


seven approx i mate l y even ly s paced levels of Bi . At each o f the ex treme
values , exper iments at a central l evel of both Ta and T in were
performed . At each of the f i ve i ntermed iate Bi levels a fac tor ial
des i gn w i th Ta and T in each at two l ev e ls plus one centrepoint was
conduc t ed and some other runs at int ermed iate l evels were made . Evenly
spaced l evels of 1 / St e , corrosponding to Ta values of ab ou t 5 °C , 1 3 °C
and 45 °C , and Pk wi th values of T i n o f abou t - 1 0 °C , - 20 °C and -30 °C
were used , with h vary i ng fr om 1 3 . 2 W m- 2 oc - 1 to 1 7 2 . 7 W m - 2 oc - 1 and
D from 0 . 026 m to 0 . 1 05 m r espec t i ve l y . Th i s des ign was cons i dered to
g i ve a sat i sfactory data base for assessment of thaw i ng t ime prediction
methods .

The resu l t s of the f u l l set of 3 5 slab thawi ng exper iments i nc lud ing
r ep l i cates to est i mate the exper imental error are shown in Tab l e 6 . 2 .
Typ i cal thaw i ng t emperature / t ime p ro f i l es are shown i n Fi g . 6 . 1 . The
rep l i cates were p er formed as independent ly from each o ther as pos s ible
to g iv e a true i nd icat i on of var iab il i ty due to experimental
t echn i qu es . The only var i able that could not be measured independent l y
from the other slab exper iments was the sur face heat transf er
coef f ic i en t . Any add i t ional uncerta inty i ntroduced by errors in
Exper imental Des i gn and Results 125

measurement , and due t o systema t i c error , cou l d not b e quan t i f ied from
the s pread of r ep l i cates .

S i x thaw i ng exper iments wer e conduc t ed w i th mi nced l ean beef for a


repres ent a t i ve range of cond i t ions to check that exper i ments wi th
Tylose gave resu l t s cons i st en t w i th those for rea l foodst u ffs , and that
N 2 and N 3 wer e not parameters s i gn i f i cant ly affec t i ng p red ict ion method
accuracy . Results for the m i nced l ean beef exper i ments are g i v en i n
Tab l e 6 . 8 and F i g . 6 . 1 7 i s a t y p i c a l t emperature/t ime p ro f i l e .

6.3 THAW I NG OF I NFINITE CYLI NDERS

For i n f i n i t e cyl inders the range of Biot numb ers that cou ld be
cons i dered was phy s i ca l ly l i mi ted by the cyl inder d i ameter s and p i pe
wal l mat er i a ls and th i ck nesses used . By use of both meta l and P VC
wal led cyl i nders and the add i t ion of rubber insulat ion a range of h
from 1 9 . 0 W m- 2 oc - 1 to 1 1 3 . 0 W m -2 oc - 1 was obtained in the l i qu id
i mmers ion system . In comb inat ion wi th d i fferent cyl inder d iameters
th i s gave s i x app rox imate l evels in B i from 1 .8 to 26 . 0 . At each l ev e l
a s im i lar des i gn t o that used for s l abs was s e t up to cover a w i de
range of amb i ent and i n i t i a l t emperature cond i t ions .

The 3 4 exper i menta l resu l t s for thawi ng inf i n i t e cyl i nders of Tyl ose
are g iven in Tab l es 6.3, and a typical thaw i ng curve is shown in
Fig. 6 . 2 .

6.4 THAW ING OF SPHER ES

Th irty-fi ve thaw i ng runs wer e conduc ted w i th Ty lose s pheres . A s i m i l ar


des ign to that used for inf ini t e cyl i nder tha w i ng was used . The metal
wall ed spheres plus
rubber coat ing gave surface heat t ransfer
coeff i c ients from 41 . 9 W m- 2 o c- 1 to 2 46 . 2 W m -2 oc- 1 . For the three
sphere d iame t ers used . there were s i x l eve l s o f the Blot number in the
range 5 . 9 t o 57 . 0 .

F i gure 6 . 3 g ives a typ i ca l thaw i ng t emper ature/ t i me pro f i l e wh i l e the


full set of result s are g iv en i n Tab l e 6 . 4 .
Exper imenta l Des i gn and Results 1 26

6.5 THAW I NG OF R ECTANGU LAR BR ICKS

The thaw i ng t ime of a rectangular b r i ck shaped object dep ends on its


geometry a s wel l a s the three parame ters Bi , S t e and Pk . The shap e o f
a rec tangular br ick i s defi ned b y t h e rat io of each of the two l o nger
s i de l eng ths to the shor test (S1 and S2) . The shortest s i d e l eng th i s
equa l t o the character i s t i c d imens ion ( D ) so the two shape f actors , S1

and S2, can b e i ntroduced . The exper iments were des igned to cover a l l
the f i ve fac tors ( B i , Ste , P k , S1 and S 2) as w i de ly as was pract i cab l e .

For each o f twelve d i fferently shaped and s i z ed boxes a 22 factor i a l


design i n S t e and P k p lus a centrepo int ( corr espond i ng to v a l ues o f T a
of 5 °C , 1 3 °C and 45 °C , and T in of - 1 0 °C , -20 °C and - 30 °C ) was
per formed . The f i ve runs for each b r i ck wer e d i v ided between two
l eve ls of the Biot numb er ( ar i s i ng from the boxes both wi th and wi thou t
rubber sheets glued to the surfac e ) . The geome t r i c fac tors , S1 and S2,

were var i ed from 1 to 4 , th i s range cover i ng p rac t i cal appl icat ions .

The results for the 68 thaw i ng exper iments conduc ted are shown i n Tabl e
6 .5. A typ i ca l thawing c urve i s shown i n Fig . 6 . 4 . Four tha w i ng runs
wer e made w i th two d i fferent rec tangular b r i ck shapes u s i ng mi nced l ean
b ee f . These r esul ts is g i ven in Tabl e 6 .8 and a typ i ca l
temperature/t ime p ro f i l es i s shown in Fig . 6 . 1 8 .

6 .6 TWO-D IM ENSIONAL IRR EGU LAR SHAPES

E ight d i f f er ent t wo - d imens ional obj ects were used ; f our were
constructed by d i stort i o n of a 0 . 1 m d i ame ter PVC p i pe , and four from a
0 . 1 5 m d iame ter PVC p i pe . S i x of the obj ec ts were irregular i n shap e ,
wh i l e two were c y l i nd r i c a l ( one for each pipe s i ze ) . Figs. 5 . 1 5 to
5 . 2 1 show the cross- sect ional geome tr i es for these obj ec ts . For e ach
shape at l east four freez i ng and four thaw i ng runs were conducted .

For freez i ng three l eve l s o f the Ste fan number ( correspond i ng to values
o f Ta of abou t - 20 °C , - 28 °C and -35 °C ) and four l e vels o f the Plank
number ( correspon d ing t o T i n values o f abou t 4 °C , 1 0 °C , 20 °C a n d 30 °C )
were use d . A compl e t e fac tor i a l des i gn was not poss i b l e . The
comb inat ions of Ste and Pk used for i nd iv i dual exper i ments wer e chosen
Exper imental Des i gn and Results 127

s o that the Ste and P k levels were not stat i s t ically corre l ated over
the full data set . For thaw ing a s i mi l ar des i gn to that used for the
regular shapes was app l i ed for T a and T in · The three values of h for
each comb inat ion o f p i pe wal l th i ckness and rubber l ed t o e leven
d i f ferent l e vels of Biot number .

A tota l of 83 exper i ments wer e per formed w i th Ty los e . The results are
g i ven in Tab l e 6 . 6 and F ig s . 6 . 5 to 6 . 1 2 shown ty p i ca l temperatur e / t i me
pro f i l es at d i fferent pos i t ions w i th in each shap e . A s well , e i ght runs
were conduc ted wi th m i nced l ean beef for two of the irregular shap ed
obj ects . Tab l e 6 . 8 g i ves the results of these exper i ments . Fig. 6 . 1 9
shows typ i ca l temeperatur e / t i me pro f i l es for the minced l ean bee f runs .

6 .7 THR EE-D IMEN S IONAL IRR EGULAR SHAPES

E i ghteen ex per i men ts were conduc ted w i th the three three-d ime ns iona l
i rregul ar Ty lose obj ec ts chosen for s t udy . The shapes ar e shown i n
F i g s . 5 . 25 t o 5 . 27 . A n add i t iona l s i x runs wer e per formed usi ng two
s pher es . The exper imenta l des ign used was s im i lar to that u sed for
two-d imensional i r regular shapes . Each shape was b oth fr ozen and
thawed over a range of amb i ent and ini t ia l temperatur es . The B i ot
number was comp l e t e ly p re-set for each shape b y the mat e r i a l and size
o f construc t ion for each obj ec t . Tab l e 6 .7 g i ves the ful l s et o f
results . F i g s . 6 . 1 3 t o 6 . 1 6 show t y p i c al t emperature/ t i me pro f i l es .
E xperimental Des i gn and Resul t s 1 28

Tabl e 6 . 1 Typ i ca l Cond i t ions in Food Freez i ng and Thawing Proc esses

Free z i ng Thaw i ng

D ( m) 0 . 00 1 to 0 . 5 0 . 0 0 1 to 0 . 5

h (W m
-2 o c -1 ) 7 to 600 2 to 1 0 000

Bi 0 . 05 to 60 0 . 0 1 t o 1 000
( o
T c) -1 5 to - 40 5 to 50
a
Ste 0 . 1 2 to 0 . 3 5 0 . 07 to 0 . 9
( o
T. c ) 0 to 40 -1 0 to - 3 5
1n
Pk 0 to 0 . 6 0 . 0 6 to 0 . 3
( o
T c ) -1 0 to -30 0 to 1 0
fin
Nl 0 . 05 to 0 . 25 0 . 0 1 to 0 . 1

N2 1 . 7 to 2 . 0 1 . 7 to 2 . 0

N3 0 . 27 to 0 . 3 5 0 . 27 to 0 . 3 5
Exper imenta l Des i gn and Resul t s 1 29

Tab l e 6 . 2 Exper i mental Data For Thaw i ng of Sl ab s o f Ty lose

Run D h T T. t
X a 1n exp
Number
-2 oc - 1 ) ( hr s )
(m) (W m ( oc ) ( oc )

T1 0 . 0260 13.2 12.8 -20 . 9 4.69


T2 0 . 02 60 24 . 5 5.2 - 29 . 4 5 . 63 *
T3 0 . 05 25 13.2 5.2 -1 1 . 4 1 9 . 32
T4 0 . 0260 2 4 .5 45 . 9 -8.3 0 . 90
T5 0 . 05 25 1 3 .2 46 . 1 -28 . 6 3 . 89
T6 0 . 0525 1 3 .2 12.8 -20 . 5 10 .33
T7 0 . 0 2 60 50 . 4 4 .6 -26 . 3 3 . 49
T8 0 . 0 525 24 .5 5.2 -1 0 .7 1 3 . 20
T9 0 . 1 000 1 3 .2 45 . 9 -8 . 3 7 . 82 *
T10 0 . 1 0 20 1 3 .2 43 . 0 -1 3 .7 8 . 83
T1 1 0 . 0770 18.2 46 . 2 -26 . 8 5. 1 1
T1 2 0 . 0 525 29 . 5 1 2.8 -20 . 9 6 . 26 *
T1 3 0 . 0770 37 . 3 5.2 -30 . 2 1 8 .49
T1 4 0 . 0525 50 . 4 5.1 -25 . 0 9. 41
T1 5 0 . 02 60 78 . 1 5.2 -1 2.3 2 .5 1 *
T1 6 0 . 1 000 24 . 5 1 2.8 -9 . 4 1 6 . 92
T17 0 . 0 525 50 . 4 46 . 0 -1 0 . 6 1 .64
T1 8 0 . 0770 37 . 3 46. 1 -2 4 . 7 3 . 72
T1 9 0 . 1 000 24.5 45.8 -32 . 5 6 .75
T20 0 . 02 60 78 . 1 12.4 - 27 . 7 1 42

T21 0 . 0525 50 . 4 1 3 .4 - 20 . 2 4.68


T22 0 . 05 25 50 . 4 1 3.4 -24 . 1 4 . 65
T23 0 . 0525 50 . 4 1 3.4 -23 . 6 4 . 60
T2 4 0 . 0525 50 . 4 13.4 -2 3 . 6 4 . 50
T25 0 . 0525 78 . 1 5.2 -28 . 9 7 .58
*
T26 0 . 1 050 37 . 3 5.2 -1 0 . 4 29 . 3 3
T27 0 . 0 770 50 . 4 45.7 -1 3 .5 2 . 99
T28 0 . 0 280 172.7 43 . 0 -31 .o 0 . 43
T29 0 . 05 25 78 . 1 1 3 .3 -22 .5 3.91 *
T30 0 . 1 050 78 . 1 5.2 - 28 . 8 23 . 62
T31 0 . 0770 78 . 1 5.0 -1 4 . 2 1 4 . 39
T32 0 . 07 70 78 . 1 45 . 7 -9 . 4 2 . 61
T33 0 . 1 000 78 . 1 46 . 2 -28 . 2 4.59
T3 4 0 . 0770 78 . 1 1 2.9 -2 1 . o 7 . 25
T35 0 . 1 0 50 1 72 . 7 1 3.4 -23 . 8 1 0 .61

*
ind i cates a run i n wh ich edge heat transfer was c a l c ulated to b e
greater than 1 . 0 % .
Experimental Des i gn and Results 1 30

Tab l e 6 . 3 Exper i mental Data For Thaw i ng of Inf i n i t e Cyl inder s of Ty lose

Run D h T T. t
r a 1n exp
Numb er
oc ( oc ) ( oc )
-2 -1
( m) (W m ) ( hr s )

C1 0. 158 23 . 5 43 . 3 -1 4 .0 7.34
C2 0 . 1 58 23 . 5 21 . 1 -20 . 6 1 2 . 43
C3 o. 1 5 8 23 . 5 5.1 -28 . 4 34 . 4 1
C4 0 . 1 56 90 . 7 43 . 2 -1 1 9
• 5.31
C5 0 . 1 56 90 . 7 1 3.0 -1 3 . 6 1 2 . 30
C6 0 . 1 56 90 . 7 5.1 -27 . 9 23 . 7 4
C7 0. 1 56 43 . 5 40 . 3 -21 . 2 6 .09
CB 0 . 1 56 43. 5 1 1 .9 -27 . 4 1 5 . 40
C9 0 . 1 56 43 . 5 8.2 -26 . 9 19 .21
C10 0 . 1 56 43.5 5.3 -1 4 . 9 25 . 70
c1 1 0 . 1 06 1 13 .0 43 . 3 -1 0 . 7 2 . 47
C1 2 0 . 1 06 1 1 3.o 8 .5 -20 . 2 7 . 94
C1 3 0 . 1 06 1 1 3 .o 5. 1 -28 . 8 1 1 . 36
C1 4 0 . 1 03 37 . 4 43 . 3 -30 . 5 3 . 26
C1 5 0 . 1 03 37 . 4 1 3 .0 -1 4 . 5 7 . 25
C1 6 0 . 1 03 37 . 4 5.3 -1 0 .6 1 3 . 89
C17 0 . 1 03 25 . 1 40 . 3 -1 0 . 6 3 .7 9
C18 0 . 1 03 25 . 1 1 8 .7 -1 4. 1 6 . 63
C1 9 0 . 1 03 25 . 1 1 3.2 -1 4 . 4 8 .84
C 20 0 . 1 03 25 . 1 5.3 -31 . 2 1 6 . 73
C21 0 . 1 03 1 9 .5 43 . 9 -26 . 9 4 . 33
C22 0 . 1 03 19.5 1 8.3 -1 4.9 7 . 54
C23 0 . 1 03 1 9 .5 5.8 -1 3. 1 1 8 . 20
C2 4 0 .05 1 46 . 5 44 . 0 -1 0 .6 0 . 87
C25 0 . 051 46 . 5 8 .5 -1 1 9
• 2 .93
C26 0 . 05 1 46 . 5 5.1 -10.0 4 . 22
C27 0 . 05 1 27 . 9 40 . 3 -1 1 8
• 1 . 30
C28 0 . 05 1 27 . 9 18.9 - 26 . 5 2 . 49
C29 0 . 05 1 27 . 9 1 3 .2 -1 8 .5 3 . 00
C30 0 . 05 1 27 . 9 5.3 -28 . 0 5 . 96
C31 0 . 05 1 1 9 .0 43 . 9 - 28 . 1 1 .64
C32 0 . 05 1 1 9 .0 1 4 .6 -18.2 3 . 76
C33 0 . 05 1 1 9 .0 9 .6 -28 . 2 5 .03
C34 0 . 05 1 19.0 5 .8 -1 2. 1 7. 15
Exper imental D e s i gn and Resu lts 131

Tab l e 6 . 4 Exper i ment a l Data For Tha w i ng of Spheres of Ty l ose

Run D h T T. t
r a 1n exp
Number
-2 -1 ( hr s )
(m) (W m oc ) ( oc ) ( oc )

S1 0 . 1 28 246 . 2 43 . 3 -9.7 2 . 26
S2 0 . 1 28 246 . 2 21 . 1 -25 . 3 3 . 92
S3 0 . 1 28 246 . 2 1 3 .0 -1 9 .7 5 . 42
S4 0 . 1 28 74.8 44 . 0 -15. 1 2 . 66
S5 0 . 1 28 74.8 18.3 -1 8 . 8 4 . 78
S6 0 . 1 28 74.8 1 1 .9 - 23 . 1 6 . 39
S7 0 . 1 28 74.8 5.3 -26 . 5 1 1 . 33
S8 0 . 1 28 51 .6 43 . 3 -27 . 9 2 . 76
S9 0 . 1 28 51 . 6 22 . 0 -1 8 . 8 4. 31
S10 0 . 1 28 5 1 .6 1 4 .5 -20 . 3 5 . 81
s1 1 0 . 1 28 51 .6 5.1 -1 5 . 5 1 2 . 26
S1 2 0 . 1 28 41 . 9 43 . 6 -20 . 3 3. 13
S13 0 . 1 28 41 . 9 1 2 .0 -33 . 0 7 . 63
S1 4 0 . 1 28 41 .9 5 .5 -1 7 . 1 1 2 . 82
S1 5 0.1 12 76 . 0 43 . 9 -1 4.5 1 . 89
S1 6 0.1 12 76 . 0 8.0 -13 .9 5 . 96
S17 0. 1 1 2 76 . 0 5.3 -28 .0 7 . 87
S18 0.112 59 . 4 43 . 6 -27 . 1 2 .02
S1 9 0.1 12 59 . 4 1 4 .5 -1 6 . 6 4 .05
S20 0.112 59 . 4 5 .0 -18.2 9 . 02
S21 0. 1 1 2 45 . 7 43 . 6 -30 . 2 2 . 43
S22 0.1 12 45 . 7 8.9 -32 . 9 7. 18
S23 0. 1 1 2 45 .7 5.5 -1 7 . 4 9 . 67
S24 0 .056 1 37 . 2 18.3 -1 6 . 9 0 . 86
S25 0 . 05 6 1 37 . 2 1 1 .9 -1 3 . 4 1 . 15
S26 0 .056 1 37 . 2 5 .3 -2 3 . 5 2 . 08
S27 0 . 05 6 87 . 0 43 . 6 -1 4 . 2 0 .58
S 28 0 .056 87 . o 22. 2 -20 . 3 0 . 87
S29 0 . 05 6 87 . 0 1 4.5 -1 4 . 8 1 11•

S30 0 .056 87 . 0 7.0 -2 2 . 8 1 . 88


S31 0 . 056 87 . 0 5.0 -24 . 7 2 . 33
S32 0 .056 57 . 5 43 . 6 -22 . 8 0 . 67
S33 0 . 056 57 . 5 22 . 3 -1 4 . 9 1 06•

S3 4 0 . 05 6 57 . 5 12.1 -21 . 7 1 . 58
S35 0 . 0 56 57.5 5.5 -1 6 . 3 2 . 67
Exper i mental Des i gn and Results 132

Tabl e 6 . 5 Exper i mental Data For Thaw i ng of Rectangular Br i ck s of Tyl ose

Run D D D h T T. t
X y z a 1n exp
Numb er
oc ( oc ) ( oc )
(m) -2 -1
( m) (m) (W m ) ( hrs )

81 0 . 075 0 . 075 0 .075 41 . 0 43 . 2 -28 . 5 1 . 67


82 0 . 0 75 0 . 0 75 0 . 075 41 . 0 5.2 -1 1 .9 6 � 86
83 0 . 075 0 . 075 0 . 075 23 . 7 39 . 1 -1 0 . 5 2 . 05
84 0 . 0 75 0 . 0 75 0 . 075 23 . 7 2 1 .5 -25 . 3 3 . 22
85 0 . 07 5 0 . 075 0 . 0 75 23 . 7 l5 . 1 - 23 . 7 4.07
86 0 . 0 75 0 . 075 0 . 0 75 23 . 7 5.8 -29 . 9 8 . 07
87 o. 104 0 . 1 25 0 . 1 52 41 . 0 45. 0 -1 2 . 0 3 . 50
88 0 . 1 04 0 . 1 25 0 . 1 52 41 . 0 1 3.8 -23 . 3 8 . 35
89 0 . 1 04 0 . 1 25 0. 152 41 . 0 5.2 -26 . 5 1 5 . 98
81 0 0 . 1 04 0 . 1 25 0 . 1 52 23 . 7 39. 1 -30 . 8 4 . 92
B1 1 0 . 1 04 0 . 1 25 o. 152 23 .7 5.7 -1 2 . 4 1 7 .65
81 2 0 . 05 2 0 . 078 0 . 20 2 41 .0 1 3.8 -22 .5 3 . 45
81 3 0 . 052 0 . 0 78 0 . 202 41 .0 5.3 -26 . 4 6 . 88
81 4 0 . 05 2 0 . 078 0 . 20 2 41 . 0 5.2 -1 2 . 4 6 . 84
81 5 0 . 0 52 0 . 07 8 0 . 20 2 23 . 7 39 . 1 -1 1 . 6 2 . 05
81 6 0 . 05 2 0 . 07 8 0 . 202 23.7 21 .5 -3 2 . 0 3 . 42
81 7 0 . 081 0. 1 5 1 0 . 1 51 41 . 0 46 . 1 -31 . o 3.16
81 8 0 . 08 1 0. 151 0. 151 41 .0 1 3.8 -21 . 8 7 . 09
81 9 0 .081 0. 1 51 0. 1 5 1 41 . 0 5.2 - 29 . 4 1 3 . 49
820 0 . 08 1 0. 151 0.151 23 . 7 39 . 1 -1 0 .7 4.10
82 1 0 .081 0. 1 5 1 0 . 1 51 23 .7 5.7 -1 2 . 9 1 5.81
822 0 . 054 0 . 1 27 0 . 20 1 41 .0 45 . 6 -1 2 . 1 1 .84
823 0 .054 o . 1 27 0 . 20 1 41 . 0 5.2 - 28 . 3 8 .59
82 4 0 . 0 54 0 . 1 27 0 .201 23 . 7 21 .6 -18 .7 4 . 45
825 0 .054 0 . 1 27 0 . 20 1 23 .7 15. 1 -1 7 .7 5 . 72
826 0 .054 0 . 1 27 0 . 20 1 23 . 7 5 .7 -10 .5 1 0 .66
827 0 . 0 60 0 . 20 1 0 . 20 1 41 . 0 5.3 -32 . 0 1 1 92

828 0 . 0 60 0 . 20 1 0 . 20 1 41 . 0 5.2 -31 . 0 1 1 . 50


829 0 . 060 0 . 20 1 0 . 20 1 23 . 7 43 . 3 -1 1 . 9 3.1 3
83 0 0 .0 60 0 . 20 1 0 . 20 1 23 . 7 21 .5 -1 6 . 9 5 . 46
831 0 . 0 60 0 . 20 1 0 . 20 1 23 . 7 1 3 .9 -1 7 . 2 7 . 44
832 0 . 0 60 0 . 20 1 0 . 20 1 23 . 7 5 .7 -1 1 . 8 1 3 . 82
833 0 . 1 09 0 . 1 56 0 . 207 28 1 . o 42 . 9 - 30 . 0 3 . 60
83 4 0 . 1 09 0 . 1 56 0 . 20 7 281 . 0 24 . 5 -1 1 . 8 5 . 29
835 0 . 1 09 0 . 1 56 0 . 207 79 . 1 1 3 .2 - 23 . 8 9 . 63
836 0 . 1 09 0 . 1 56 0 . 207 79 . 1 10.0 - 27 . 9 1 1 . 80
837 0 . 1 09 0 . 1 56 0 . 207 79 . 1 5.4 -1 2 . 1 1 7 .00
83 8 0 . 1 58 0 . 1 59 . o . 1 64 28 1 . 0 42 . 9 -1 2 . 0 4 . 53
83 9 0 . 1 58 0 . 1 59 0 . 1 64 28 1 . 0 1 2 .8 -27 . 9 1 0 .5 4
840 0 . 1 58 0 . 1 59 0 . 1 64 28 1 .0 5.6 -31 . 4 1 9 .05
841 0 . 1 58 o . 1 59 0 . 1 64 79 . 1 41 . 3 -30 .0 5 . 75
842 o . 1 58 o . 1 59 0 . 1 64 79. 1 10.0 -21 . 7 1 4 . 70
843 0 . 1 5 8 0 . 1 59 o. 164 79.1 5.4 -1 1 . 6 20 . 8 6

• • • cont inued
Tab l e 6 . 5 cont inued • • •
1 33

Run D D D h T T. t
X y z a 1n exp
Number -2 -1
( m) (m) ( m) (W m oc ) ( oc ) ( oc ) ( hr s )

844 0 . 08 1 0 . 093 0 . 1 96 28 1 . 0 42 . 9 -9 .5 1 .65


845 0 . 08 1 0 . 093 0. 1 9 6 281 . 0 20 . 4 -1 2 . 0 2 . 89
846 0 . 081 0 .093 0 . 1 96 28 1 . 0 9 .3 -1 1 . 0 5 .52
847 0 . 08 1 0 . 093 0 . 1 96 281 . 0 5 .6 - 1 0 .6 7 . 40
848 0 . 081 0 .093 0 . 1 96 79. 1 41 . 3 -29 . 2 2 . 35
84 9 0 . 08 1 0 . 0 93 0 . 1 96 79 . 1 1 3 .2 -1 2 . 1 5 . 07
850 0 .081 0 . 093 0. 1 96 79 . 1 5.4 -29 . 2 9 . 25
85 1 0 . 082 0 . 1 54 0 .233 28 1 . 0 42.9 - 29 . 5 2 . 49
852 0 . 082 0 . 1 54 0 . 233 28 1 .0 12.8 -1 3 .0 5 . 97
85 3 0 . 082 0 . 1 54 0 . 23 3 28 1 . 0 5 .6 -8 .5 1 0 . 39
854 0 . 082 0. 1 5 4 0 . 233 28 1 . 0 5.1 -1 0.9 1 1 39

855 0 . 082 0 . 1 54 0 . 233 79 . 1 41 .3 -1 1 . 4 3 .05


856 0 . 0 82 0 . 1 54 0 . 23 3 79.1 20 . 7 -22 . 8 5.18
85 7 0 . 082 0 . 1 54 0 .233 79 . 1 5.4 -19 .7 1 2 . 79
858 0 . 079 0 . 079 0. 1 56 28 1 . 0 43 . 3 -29 . 0 1 . 55
859 0 . 07 9 0 . 0 79 0 . 1 56 281 . 0 9.3 -10.0 4 . 39
860 0 . 079 0 . 07 9 0 . 1 56 28 1 . 0 5.6 -22 . 0 6 . 02
861 0 . 0 79 0 . 07 9 0. 1 56 79 . 1 20 . 7 -1 1 . 0 2 . 97
862 0 .079 0 . 07 9 o. 1 56 79 . 1 1 3.2 -21 . 5 4 . 29
863 0 . 079 0 . 07 9 o . 1 56 79 . 1 5.4 - 28 . 9 7 . 92
864 0 . 0 86 0 . 3 03 0 . 306 28 1 . 0 1 4.6 -22 . 7 6 . 44
865 0 . 086 0 . 303 0 . 30 6 28 1 . 0 5 .6 -1 1 5 • 1 1 93

866 0 . 086 0 . 30 3 0 . 306 79 . 1 41 .3 -1 0 . 3 3 . 45


867 0 . 086 0 . 30 3 0 .306 79 . 1 1 3 .2 -2 3 . 5 8 . 43
868 0 . 0 86 0 . 30 3 0 . 306 79 . 1 5.4 -27 . 4 1 5 .99
Exp e r i menta l D es i gn and Resul t s 1 34

Tab l e 6 . 6 Exper i menta l Data For Freez ing and Thawing of Two- D i mens i onal
I r r egular Shapes of Ty lose

Run Shap e D h T T. t
a 1n ex p
Number Code
-2 oc - 1 ( oc ) ( hr s )
(m) (W m ) ( oc )

11 o. 1 5 25 28 . 0 21 . 1 -1 5 . 2 9 .95
12 0 . 1 5 25 28 . 0 20 . 8 -1 6 . 6 1 0 . 60
13 0 . 1 525 28 . 0 1 3 .7 -1 3 . 7 1 3 . 79
14 o . 1 5 25 28 . 0 13 6 • -17 . 2 16.1 2
15 0 . 1 5 25 28 . 0 5.6 -1 6.1 26 . 23
16 0 . 1 5 25 28 . 0 - .7
1 9 18.1 9 . 83
17 0 . 1 5 25 28 . 0 -23 . 6 2.8 6 .77
18 0 . 1 5 25 28 . 0 -25 . 0 33.6 7 . 88
19 0 . 1 52 5 28 . o -27 . 0 19 .4 7.34
110 0 . 1 5 25 28 . 0 -30 . 6 2 .7 5 . 82
11 1 0 . 1 525 28 . 0 -33 . 9 19 .4 5 . 45
112 0 . 1 5 25 28 . 0 -3 8 . 5 23.9 5. 18
113 2 0.1 1 15 28 . 0 40 . 6 -1 2 . 1 5 . 40
11 4 2 0.11 15 28 . 0 20 . 3 -31 . 3 9 . 98
115 2 0. 1 1 15 28 . o 1 3.7 -21 . 9 1 3 .28
116 2 0.1 1 15 28 . 0 1 3 .6 -1 7 .2 13. 1 1
117 2 0.1 1 15 28 . 0 5.6 -31 7• 24 . 9 8
118 2 0. 1 1 15 28 . 0 -19 .7 31 .8 9 .53
119 2 0.1 1 15 28 . 0 -23 . 7 4.3 6.31
1 20 2 0. 1 1 15 28 . 0 - 25 . 1 34.9 7 . 30
121 2 0.1 1 15 28 . 0 -26 . 9 20 . 0 6 .70
122 2 0. 1 1 15 28. 0 - 38 . 6 1 .8 3 . 96
1 23 3 0 . 1 370 28 . 0 20 . 3 -1 1 .2 10. 17
124 3 0 . 1 370 28 . 0 1 3 .7 -19.9 1 4 . 47
1 25 3 0 . 1 370 28 . 0 8.4 -1 9 . 4 1 8 .69
126 3 0 . 1 370 28 . 0 5 .6 -18.4 27 . 32
1 27 3 0 . 1 3 70 28 . 0 -1 9 .8 1 9 .5 9 .31
1 28 3 0 . 1 370 28 . 0 -26 . 9 20 . 1 7. 13
1 29 3 0 . 1 370 28 . 0 -30 . 5 32 . 8 6 . 90
130 3 0 . 1 370 28 . 0 -38 . 5 13.4 4 . 63
131 4 0 . 0 9 90 28 . 0 20 . 1 -21 . 5 8 .98
132 4 0 . 0990 28 . 0 1 3 .7 -3 2 . 5 1 2 . 28
1 33 4 0 . 0 990 28 . 0 8.4 -1 3 . 2 1 6 . 86
134 4 0 . 0 990 28 . 0 5 .6 -9 . 5 21 1 9 •

1 35 4 0 . 0 9 90 28 . 0 -1 9 . 8 18.1 7 .82
136 4 0 . 0990 28 . 0 -27 . 1 1 9 .9 6.17
137 4 0 . 0 990 28 . 0 -38 . 6 19.1 4 .0 4

• • • cont inu e d
Tab l e 6 . 6 cont inued • • •
1 35

Run Shape D h T T. t
a 1n exp
(o (o
Number Code
-2 oc- 1 c c)
( m) (W m ) ) ( hr s )

1 38 5 0 . 1 050 34 . 2 21 . 1 -1 5 . 7 5.1 3
1 39 5 0 . 1 0 50 34.2 13 .6 -1 2.7 7.19
1 40 5 0 . 1 050 34.2 8.5 -1 2 . 6 9 .35
141 5 0 . 1 0 50 34 . 2 5 .6 -32 . 5 1 5 . 17
1 42 5 0 . 1 050 34 . 2 -1 9 . 8 3.3 4.51
1 43 5 0 . 1 0 50 34 . 2 -27 . 2 1 .3
9 3 . 78
1 44 5 0 . 1 0 50 34 . 2 -30 . 6 3 .1 3.01
145 5 0 . 1 0 50 34 . 2 - 39 . o 32. 4 2.71
1 46 5 0 . 1 0 50 20 . 3 5.3 -28 . 1 1 8 . 83
1 47 5 0 . 1 0 50 20 . 3 -2 2 . 2 1 4 .9 6 . 27
1 48 5 o . 1 0 50 20 . 3 -34 . 1 1 9 .7 4.1 2
149 6 0 . 0850 34 . 2 40 . 4 -29 . 8 3.19
150 6 0 . 0 85 0 34. 2 20 . 2 -20 . 1 5 . 23
15 1 6 0 . 0850 34.2 1 3 .7 -20 . 9 7 14

1 52 6 0 . 0 850 34.2 5.5 -9 . 7 1 3 .08


1 53 6 0 . 0850 34 . 2 -19 .6 31 . 4 5 . 01
154 6 0 . 0 850 34 . 2 -26 . 9 20 . o 3 .5 5
155 6 0 . 0850 34 . 2 -30 . 6 22. 0 3.21
1 56 6 0 . 0 85 0 34.2 -39 . 1 2.3 2 .05
157 7 0 . 09 1 0 34 . 2 . 20 . 2 -31 . 6 5 . 24
158 7 0.091 0 34 . 2 13.7 -21 . 4 7 . 32
1 59 7 0 . 09 1 0 34 . 2 8.4 -29 . 4 9 . 06
1 60 7 0 .091 0 34 . 2 5.6 -28 . 4 1 3 . 60
161 7 0 . 09 1 0 34 . 2 -19 .8 2 .6 4 . 30
1 62 7 0.091 0 34.2 -26 . 9 20 . 6 3 . 82
163 7 0 . 09 1 0 34 . 2 -30 . 1 1 2.8 3.17
1 64 7 0.091 0 34 . 2 -38 . 8 28 . 4 2 .58
1 65 7 0 . 09 1 0 20 . 3 32 .5 - 28 . 7 4 . 55
1 66 7 0.0910 20 . 3 8.1 - 26 . 6 12.18
1 67 7 0 . 09 1 0 20 . 3 5 .3 -1 1 .6 1 6 .8 1
1 68 7 0 .091 0 20 . 3 -21 . 3 1 9 .9 6 .5 6
169 7 0 . 09 1 0 20 . 3 - 29 . 9 3.3 3 . 77
1 70 7 0 .09 1 0 20 . 3 -3 4 . 1 20 . 4 3 .87
171 8 0 . 0475 34 . 2 40 . 6 -1 1 9
• 1 . 76
1 72 8 0 . 0 4 75 34.2 20 . 6 -1 1 . 4 3 .01
173 8 0 . 0 475 34 . 2 1 3 .7 -20 . 2 4 . 43
174 8 0 . 0 4 75 34 . 2 5.6 -1 5 . 1 7 .99
1 75 8 0 . 0 475 34 . 2 -1 9 .7 14.4 3 . 40
176 8 0 . 0 4 75 34 . 2 -27 . 1 19.4 2 . 65
177 8 0 . 0475 34.2 -38 . 7 1 8 .3 1 .70
178 8 0 . 0 4 75 20 . 3 1 5.6 -1 2 . 1 4 . 70
179 8 0 . 0475 20 . 3 8. 1 -1 1 .2 7 . 69
1 80 8 0 . 0 475 20 . 3 5.3 -27 . 6 1 1 . 77
181 8 0 . 0 475 20 . 3 - 22 . 1 2.9 3 . 87
1 82 8 0 . 0 475 20 . 3 -29 . 6 33 . 7 3 . 48
1 83 8 0 . 0 4 75 20 . 3 -34 . 1 34 . 5 3.11
Experi mental Des ign and R esu lts 1 36

Tab l e 6 . 7 E xperimental Data For Fr eez ing and Thawing of


Three-D imens iona l Irregular Shapes of Tylose

Run Shape D h T T. t
a ln exp
Number
-2 -1
( m) (W m oc ) ( oc ) ( oc ) ( hr s )

TI1 Pyrami d 0 . 0825 41 . 0 21 . 0 -1 3 . 5 4.52


TI2 Pyramid 0 . 0825 41 .0 9 .8 -1 5 .6 7 . 65
TI 3 Pyramid 0 . 0 82 5 41 . 0 5.6 -28 . 2 1 2 . 28
TI 4 Pyram i d 0 . 0825 41 . 0 -25 . 6 35.8 3 . 62
TI5 Pyramid 0 . 0 825 41 . 0 -29 . 2 5.1 2 .70
TI6 Pyramid 0 . 0825 41 .0 -38 . 0 19 .8 2 . 48
TI7 Sphere 0 . 1 270 51 . 4 12.8 -1 1 9
• 5 . 94
TI8 Sphere 0 . 1 2 70 51 .4 5 .6 -20 . 9 11 .14
TI 9 Sphere 0 . 1 270 51 . 4 -21 . 1 20 . 5 3 . 03
TI 1 0 Sphere 0 . 1 2 70 51 .4 -2 2 . 5 36 . 7 3 .55
TI 1 1 Spher e 0 . 1 270 51 . 4 -37 . 9 4 .0 1 . 83
TI1 2 Sphere 0 . 1 2 70 51 .4 -3 8 . 1 30 .7 1 . 99
TI13 Egg 0 . 1 70 0 51 . 4 30 . 1 -1 4 . 9 5.62
TI1 4 Egg 0 . 1 7 00 5 1 .4 1 2.8 -28 . 6 1 1 . 46
TI 1 5 Egg 0 . 1 700 51 . 4 5.7 -1 2 . 0 2 1 . 22
TI1 6 Egg 0 . 1 70 0 51 .4 -2 2 . 6 20 . 6 5 . 76
TI 1 7 Egg 0 . 1 700 51 . 4 -29 . 1 34 . 9 4 . 55
TI 1 8 Egg 0 . 1 700 51 . 4 -38 . 6 19 .6 4. 1 1
TI 1 9 F i sh 0 . 1 200 51 . 4 1 4.7 -1 3 . 0 8 . 29
T I 20 F i sh 0 . 1 200 51 .4 10.8 -1 3 .7 1 0 . 48
TI21 F i sh 0 . 1 200 51 . 4 5.6 - 28 . 4 18 .76
T I 22 F i sh 0 . 1 200 51 .4 -21 . 9 5.9 4.13
T I 23 F i sh 0 . 1 200 51 . 4 -29 . 1 34.4 3 .65
T I 24 F i sh 0 . 1 20 0 51 .4 -3 8 . 5 1 6 .3 2.81
Exper i menta l Des ign and Resu l ts 1 37

Table 6 . 8 Exper i mental Dat a For Freez ing and Thaw ing of Slab s and
Mult i -D i mens ional Shapes of M i nced Lean Beef

Run Shap e 1 D D D h T T. t
X a 1n exp
Number & Code
y z
( m) (m) ( hr s )

M1 Slab 0 . 024 13.2 5.8 -1 9 .1 8 .58


M2 Slab 0 . 0 24 50 . 4 26 . 9 -1 5 .6 0 . 84
M3 Slab 0 .047 18 .2 8 .0 -27 . 1 1 2 .07
M4 Slab 0 . 047 78 . 1 15.8 -24 . 4 2 . 85
M5 Slab 0 . 075 24 . 5 43 . 2 -29 . 0 5.21
M6 S l ab 0 . 0 75 17 2. 7 9.3 -1 6 . 9 8 . 94

M7 Br i ck 0. 152 0. 1 53 0 . 1 54 41 .0 27 . 0 -28 . 4 7 . 22
M8 Br i ck 0 . 1 52 0 . 1 53 0 . 1 54 41 .0 8.4 -1 5 .0 1 8 .01
M9 Br i ck 0 . 076 0 . 077 0 . 300 41 .o 1 3. 0 -22 . 3 5 . 23
M1 0 Br i ck 0 . 0 76 0 .077 0 . 300 41 .0 5.8 - 23 . 9 9 . 32

M1 1 2D1 3 0 . 1 370 28 . o 21 .0 -27 . 1 1 0 .07


M1 2 2D1 3 0 . 1 370 28 . 0 8. 1 -1 3 . 2 1 9 . 50
M1 3 2D1 3 0 . 1 370 28 . 0 -29 . 1 1 8 .8 6 . 27
M1 4 201 3 0 . 1 3 70 28 . 0 -30 . 2 24 . 5 6 . 33
M1 5 201 8 0 . 0 475 20 . 3 1 3.9 -1 1 6
• 5 . 05
M1 6 2D1 8 0 . 0475 20 . 3 6 .5 -1 6 . 3 8 . 75
M1 7 2D1 8 0 . 0 475 20 . 3 -23 . 3 3. 1 3 . 43
M1 8 2DI 8 0 . 0 4 75 20 . 3 -37 . 8 32 . 2 2 . 59

1 201 two-d imens ional irregular , code numbers f r om Tab l e 6 . 6 .


E xper imental De s ign and Results 1 38

10
T.

0
'8
0 Tc
-

-
� -10

! -20 Legend
A

B
.........••

-30 �------r---�--�---.--,
0

F i g . 6 . 1 A Typ ic a l Tempe rature / T i me Prof ile For Tha w i ng of Slab s o f


Ty lose . Run T2 4 , D x 0 . 0 5 25 m, h 50 . 4 W m -2 oc -1 ,
= =

Ta 1 3 . 4 °C , T in
= - 2 3 . 6 °C . A e x per imentally measured
= -

temp era tures , B - temp era tures pred ic ted by the fu l l f i n i te


elemen t me thod formu la t ion . Resu l t s for al l th e nu me r i ca l
me thods are a l l v ir tua l ly identic a l .
20

·
··
··

r.
··
··
··
· ··
··
10 ··
··
·

··
··
·

u
··

..
·
··

.e..,
·

·
·
··

·
··
··

·


-
0 0


a.

� Legend
A
-10
�--········

-20 �------�--r---��---,
0 2 4 6 8
Time (hrs)
F i g . 6 . 2 A Typ i cal Temperature/ Ti me Prof i le For Thaw i ng of In f ini te
Cyl inders of Ty lose.
Run C22 , Dr •0 . 1 03 m, h a 1 9 .5 W m - 2 oc- 1 , T a - 8 . 2 °C , =

T i n a - 1 4 . 9 °C . A - exper i me n tally me asured t emper a tu r es ,


B - t emperatures p red icted by the f i n it e d i fferenc e me thod .
Resu l t s for th e f ini t e d i f f erence me thod a nd b oth f i n i t e
eleme n t me thod formu la t io ns are a l l v ir tually iden t i c al .
Exper imental De s i gn and Resul t s 1 39

20

10 T.

u

CD 0
"-

-
0
"-
CD
a. -10
E Legend
.,!
A

-20 B
..........•

-30 �----�--�
0 2 4 6 8
Time (hrs}
Fig . 6 . 3 A Typ ic a l Tempe ra ture / Ti me Prof i le For Th aw i ng of Spher es o f
Ty los e . Run S6 , D r = 0 . 1 28 m , h 7 4 . 8 W m-2 °C - 1 ,
T a = 1 1 . 9 °C , T in = - 2 3 . 1 °C . A - exper i me n tally me asured
temper atures , B - temperatures predic ted by th e s i mp l i f ied
f in i te e lemen t method . Resu l t s for a l l the nu mer ica l method s
ar e v ir tua l ly iden t i c a l .
40

u 20

CD
"-

-
0
Legend
"-
CD A
a.
E �--········
.,! 0
c
--

-20 �------�---r---,
0 2 3
Time (hrs)
Fig . 6 . 4 A Ty p ic a l Temp erature/ Time Pr ofi le For Thaw i ng of Rec tangular
Br i ck s of Tylo se . Run B7 , D x 0 . 1 04 m , D y - 0 . 1 25 m, D z
= •

0 . 1 52 m, h 50 . 4 W m
s
- 2 -1
oc , T a 45 . 0 °C , T in
• - 1 2 . 0 °C .

A - exp er i mental ly measured temperatures , B - temperatures


pred icted by th e f in it e d ifferenc e me thod , C - t emperatures
pred i c t ed b y the s i mp l i f ie d f ini t e e lemen t me thod , T •
w
t emperature a t th e surfac e o f th e c en t re o f th e x , z f ace .
Exper i mental De s ign and Resu l t s 1 40
5

'U

-5

G)
-
� Legend
-10
� A
G)
a.
B
� -15 ...........

---
c

-20

-25
0 2 4 6 8
Time (hrs)
Fig . 6 . 5 A Typ ical Temperature / T ime Prof i le For Freez i ng or Tha w i ng o f
the Ty lose Two-Dimens ional Ir r egular Shape Numb er On e .
Run 1 7 , h = 28 .0 W m-2 °C -1 , T a = - 2 3 . 6 °C , T in = 2 . 8 °C .
A - e xpe r i mental ly measure d temperatures , B - tempe ratu res
pred ic ted by the full f i n i t e e lement me thod , C - temperatures
pred icted by th e s i mp l i f ied f i n i t e e lemen t me th od. Ca l culated
usi ng the two-d ime ns ional g r i d in Fi g . 5 . 1 5 .
20

10 Legend
'U A

G) B
...........

::s ..
-
0 0 ---
c

G) 3
a.


-10

-20 ;------.--�---,
0 5 10 15
Time (hrs)
Fig . 6 . 6 A Typ ical Tempera ture/ Ti me Prof i le For Freez i ng or Thaw i ng o f
the Ty lose Two- D i mens ional Ir r e gula r Shape Number Two .
Run 1 1 6 , h = 28 . 0 W m- 2 oc- 1 , T a = 1 3 . 6 °C , T in = - 1 7 . 2 °C .
A - expe r i ment a l ly measur ed t e mperatures , B - tempe r a tu res
pred icted by the ful l f i n i te e leme n t me thod , C - t empe ratures
pred i c t e d by th e s i mpl i f ie d f in i t e e leme nt method . Ca lcul ated
us ing the t wo-d i me ns io nal grid in Fi g . 5 . 1 6 .
Exper i me n ta l De s ign and Resu l t s 1 41

40

20 Legend
'G' A
t_.,


�--········
::::J
-
0 0 £._ __



Q.
E

-20

-40 4-------r---�--,--.
0 2 4 6 8
Time (hrs)
Fig . 6 . 7 A Ty p ical Temperature/ Ti me Prof i l e For Fr eez i ng or Thaw i ng o f
th e Ty lose two-D i mens ional Irregular Shape Numb er Three.
Run I29 , h = 28 .0 W m-2 °C -1 , Ta - 3 0 . 5 °C , T in = 32 . 8 °C .
=

A - expe r i menta l ly measured temper a tures , B - tempe ratures


pred ic ted by the fu l l f i n i te element method , C - temperatures
p r ed i c ted by th e s i mpl i f i e d f i n i te e l e men t meth od . Ca lcul ated
using the two-d imens ional g r i d in Fig . 5 . 1 7 .
20

10
Legend
'G'
0 A
......_,

a» 0 8
� ..........•

::::J
- c
0 ---


Q. -10


-20

-30
;---
0 2 4 6 8 10 12--.----r---r--�---,--�
Time (hrs)
F i g . 6 . 8 A Typ ical Temp erature/ T ime Prof i le For Freez i ng or Thaw i ng o f
the Ty los e Two-D i me ns io nal Ir r e gular Shape Number Fou r .
Run I 3 1 , h 28 .0 W m- 2 °C -1 , T a = 20 . 1 °C , T in = - 2 1 . 5 °C .

A - exper imenta l ly measur ed t emp era tures , B - t emperatu res


pred icted by the full f in it e e leme n t me th od , C - t empera tures
pred i c t ed by the s i mpl i f i e d f ini t e e lemen t method. Ca lcu l at e d
us ing th e t wo-d i mens ional grid i n F i g . 5 . 1 8 .
Expe r i me nta l Des ign and Resu lts 1 42
20

10

(]'
0
-

G)

-
3
0 Legend
e
G) A

! -10
B
...........

c
---

-20 ;-------�--r--�
0 2 4 6 8
Time (hrs)
Fig . 6 . 9 A Ty pical Temperature /Ti me Pr ofi le For Fr eez ing or Thaw i ng of
the Ty lo se Two- Di mens ional Irregular Sh ape Numb er Five .
Run 139 , h = 3 4 . 2 W m-2 °C -1 , T a 1 3 . 6 °C , T in = - 1 2 . 7 °C .
=

A - ex p er i men ta l ly measured tempe ratures , B - t empe ratu res


pred icted by th e full f i n i te e leme n t me thod , C - temperatures
pred i c ted by th e s i mpl i f i e d f i ni te e l emen t me thod. Ca lcula ted
using the two-d imens ional g r i d in Fi g . 5 . 1 5 .
10

0
Legend
0"
0 A
._

G) -10 B
� ...........
:::J
- c
0 ---

G)
a. -20
E
t!!
-30

-40 1-----r--r--T---�--�
0 0.5 1 1.5 2 2.5
Time (hrs)
Fi g . 6 . 1 0 A Typ ical Temp erature/ Ti me Pro f i le For Fr eez ing or Thawing of
the Ty l ose Two-D imens ional Irregular Shape Number S i x .
Run 156 , h • 3 4 . 2 W m -2 oc- 1 , T a - 39 . 1 °C , T in = 2 . 3 °C .
=

A - ex pe r i mental ly measure d temp er a tures , B - t empe ratures


pre d ic t ed by th e full f i n i te e leme n t me tho d , C - t empe ratures
p r ed ic ted by the s i mp l i f ie d f in i te e leme nt method .
Calc ulat e d us ing th e t wo-d i me ns ional g r i d i n Fi g . 5 . 1 9 .
Exper imen tal De s ign and Resu l t s 1 43

10

?

0

- Legend

Q) A.
-5

! �--········

-10

-15 �------�--r--,�---,--�
0 5 10 15 20 25
Time (hrs)

Fig . 6 . 1 1 A Typ ical Temp erature/ Ti me Prof i le For Freez i ng or Th aw i ng o f


the Ty lose Two- D i mens ional Ir regular Shap e Number Sev en .
Run 129 , h = 20 . 3 W m-2 °C - 1 , T a = 5 . 3 °C , T in = - 1 1 . 6 oc .
A - expe r i menta l ly measured tempera tures , 8 - tempera tures
pred icted by th e fu l l f in i t e e leme n t method . Ca lc ulated
us i ng the two-d i mens iona l gr i d in Fig . 5 . 20 .
20

U'
0
0
..._,


- Legend
e
G A.

! - 20
9
..........•

-40 4---------�--r--,�--�
0 0.5 1 1.5 2
rme (hrs)
Fig . 6 . 1 2 A Typ ic a l Temp erature /Ti me Pr of i le For Fre ez ing or Thawing of
the Ty l ose Two- Di mens ional I r r egular Shape Number E i ght .
Run I77 , h = 3 4 . 2 W m -2 °C - 1 , T a - 3 8 . 7 °C , T in = 1 8 . 3 °C .
s

A - exper i men t a l ly measur e d t emperatures , 8 - tempe ratures


pred icted by the full f in it e e lement me thod , C - t emperatures
pred icted b y th e s i mpl i f i e d f in i t e e lemen t me thod .
Calculated u s ing the t wo-d i men s io nal g r i d i n Fi g . 5 . 21 .
Exper imenta l De s i gn and R esu l ts 1 44
10

0
0'
0
'--"

Q)
L.

-
-10 Legend
e A.
Q)
Q.
E B
.....•...••

� c
--
-20

-30 4------r--�---�---r---.
0 0.5 1.5 2 2.5 3
Time (hrs)
Fig . 6 . 1 3 A Ty pical Te mp e rature /Time Prof i le For Fr eez ing or Thawing o f
the Ty l ose Thr ee- Di mens ional Irregu l ar Pyramid Shap e .
Run TI5 , h = 4 1 . 0 W m -2 °C -1 , T a = - 29 . 2 °C , T in = 5 . 1 °C .
A - expe r i mental ly measured temperatures , B - temperatures
pred ic ted by the fu l l f i n i t e e lement me thod , C - temp e ratures
pred i c t ed by the s impl i f ie d f i ni te e l emen t method .
Calc ulated us ing th e th ree-d i mens ional g r i d i n Fi g . 5 . 2 4 .
- at th e c entre o f the x, y fac e ( base) o f th e py rami d ,
2 - 0 . 07 65 m ( mi d-h e igh t ) above pos i t i o n 1 ( F i g . 5 . 24 ) .
15

10

0'
0 5
'--"

Q)
L.

-
c 0 Legend
L.
Q) A.
Q.
B
� -5 ...........

-10

-15 4-------r---�--�---,
0 2 4 6 8
Time (hrs)
F i g . 6 . 1 4 A Ty p i cal Tempera ture/ Ti me Prof i le For Freez i ng or Tha w i ng o f
the Ty lose Thr ee-Di mens ional Ir r e gular Sph ere Shap e .
Run TI 5 , h 5 1 . 4 W m- 2 °C -1 , T a
• 1 2 . 8 °C , T in = - 1 1 . 9 °C .

A - experi men ta l ly measure d ·temperatures , B - t empe ratures


pred icted by th e full f in i t e e leme n t me thod , C - t emperatures
pred i c te d b y t he s i mp l i f ie d f ini t e e lemen t me thod .
Calculated u s ing th e two-d imens io na l grid i n Fi g . 5 . 25 b .
Exper imen tal Des ign and Resu l t s 1 45

15

10 ·· ··
··

Legend
V'

5 A

G) B
'- .

•...•...•••

� •
- c
0 --
E
G)

! -5

-10

-15
0 5 10 15
Time (hrs)

F i g . 6 . 1 5 A Ty p ical Temperature / Ti me Prof i le For Freez i ng or Th aw i ng o f


the Ty lose Thr ee-Di mens ional Ir regu lar Fi sh Shap e .
Run TI 20 , h = 5 1 . 4 W m- 2 oc - 1 , T a = 1 0 . 8 °C , T i n = - 1 3 . 7 °C .
A - experi menta l ly measur ed tempera tur es , 8 - temperatures
pred ic ted by the full f in i t e e lement me thod , C - t emperatures
pred i c ted by the s i mpl i f i e d f ini t e e l emen t method .
Ca lc ulated us ing the three-d i me ns ional g r i d i n Fi g . 5 . 27 .
4 - 0 . 2 1 3 m a l ong th e x ax i s from th e t i p o f th e head shown
in Fi g . 5 . 27 b , 5 - 0 . 08 5 m a lo ng th e x ax i s from th e t i p of
th e head shown in Fig . 5 . 27 b , 6 - 0 . 0 2 1 m a l ong th e x ax is
from the tip of the head shown i n Fi g . 5 . 27 b and o n th e f i sh
sur fac e 0 . 021 m in th e y d irect ion from th e x ax i s .
Exper i mental De s ign and Re su l t s 1 46

30

20

Legend
'0 A

10
B
...........

- c
--
0 0
...
GJ

! -10

-20

-30
0 2 3 .. 5 6 7
Time (hrs}
Fig . 6 . 1 6 A TY p ical Te mpera ture/ T ime Prof i le For Freez i ng or Thaw i ng o f
the Ty lose Thr e e-D i me ns ional Irr e gula r Egg Shap e .
Run TI 1 6 , h = 5 1 . 4 W m - 2 °C -1 , T a = - 2 2 . 6 °C , T in = 20 . 6 °C .
A - ex pe r i ment a l ly measured tempe ratures , 8 - temperatures
pred icted by the fu l l f i n i te e lement me thod , C - tempe ratures
pred ic ted b y th e s i mp l i f ie d f i n i t e e l eme nt method .
Ca l c u lated us ing the th ree-d i mens ional g r i d i n Fi g . 5 . 26 b .

20

Legend
A

�--········

-30 4-----�--��--�-----,
0 0. 5 1 1.5 2 2.5 3 3.5
Time (lvs)
Fig . 6 . 1 7 A TY p i ca l Temp erature/ Ti me Prof i l e For Thaw i ng of Sl ab s o f
Minced Le an Bee f .
Run M4 , D x = 0 . 047 m, h 7 8 . 1 W m -2 °C -1 , T a
= 1 5 . 8 °C ,
=

Tin - 2 4 . 2 °C . A - ex per iment a l l y measured temperatures ,


=

B - t emperatures predicted by the f in i t e d if fe r enc e me thod .


Exper i me n tal De s ign and Resu l t s 1 47
10

-15 �----.---r--,
0 5 10 15 20
Time (hrs)
Fig . 6 . 1 8 A Ty p i ca l Temp e r a t u r e / T i me Pr o f i l e For Thaw i ng of Rect a ngu l ar
Br i c k s o f M i n c ed Le an B e e f .

Run M8 , D x = 0 . 1 52 m, D = 0 . 1 5 3 m, D z = 0 . 1 54 m ,
y
-2 -1
h 41 .0 W m oc , T
a
= 8 . 4 °C , T
in
= - 1 5 . 0 °C .
A - e xp e r i me n ta l l y mea s u r e d t e m pe r a t u r e s , B - t e mp e r a t u r e s

p r ed i c t e d by the f i n i t e d i f f e r e nc e me t h o d .

40

20 Legend
u
0
A.

B
...........
e
::J
- 0 c
0


a.


-20

- 40 4-----�--�---,
0 2 4 8 8 10
Time (hrs}
Fig . 6 . 1 9 A Typ i ca l Temperature / Ti me Prof i le For Freez i ng or Tha w i ng o f
a M i nc ed Le an Beef Two- D i me ns ional Irregular Shap e .
Run M1 4 , t wo-d i mensiona l irregular Shape No . 3 ,
h 28 .0 W m- 2 °C -l , T a - 30 . 2 °C , T in 24 .5 °C .
= =

A - ex per ime n ta l ly measured temperatures , B - temp eratures


pred icted by the ful l f i ni t e e lemen t me thod , C - temperatures
pred icted by the s i mp l i f ied f in i t e e leme n t me thod .
Ca lcu la t ed us i ng th e t wo-di mens iona l gr i d in Fig . 5 . 1 7 .
1 48

7 PR ED ICTION OF THAWING T IMES FOR SLA BS ,


INFINITE CYLINDERS AND SPHER ES

I n Chap . 3 pred i c t ion me thods were c l ass i f i ed into those requ i r i ng on l y


hand calculat ion ( Gr o up I ) and those requ ir ing computer calculat ion
( Gr oup I I ) . The th i r d obj ec t i ve for the p r esen t work se t in Chap . 3

was to assess the accuracy of Group I and Group II me thods for


p red i c t ion of thaw i ng t i mes for regu l ar shapes . Chap ter 7 r ep or ts
resu lts for Gro up I and Group I I me thods ap p l i e d to slab s , i n f in i te
cyl inders and spheres . Chap t er 8 repor ts the r esults o f Group I I
methods ap p l i ed to rec tangu lar b r i ck s as we l l as mu l t i -d imens ional
irregu lar shapes , wh i ls t Chap . 9 cons i ders Group I me thod s f or
mul t i -d imens ional regular shapes .

7.1 VER IFIC AT ION OF A U NIFIED AP PRO ACH FOR SIM P LE SHAPES

Pl ank ( 1 91 3 ) der i ved an equ a t i on for the t i me to freez e or thaw slabs ,


i n f i n i t e c y l inders and spheres . Th is equa t ion sugges t s that the rat io
o f t i mes f or phase change for the three shapes under iden t i ca l
cond i t ions and w i th the same charac ter i s t i c d i me ns ion i s 6:3:2. The
r a t i o ar i se s from cons i dera t ion of the relat ionsh i p be tween vo lume and
sur face area for each shape . However the equat ion requ ires that the
fol low i ng cond i t ions a r e met :
( a ) that phase change o ccurs at a un ique phase change t empera ture
( b ) that the rmal proper t i es are cons tant and
( c ) that sens i b l e heat effec t s ar e neg l ig i b l e compared w i th the l atent
heat .

I n freez ing or thawing o f b io log ical mater i a ls no t a l l these cond i t ions


ar e me t so the 6:3:2 r a t i o i s not nec essari l y cor r ec t . Howe v er tak i ng
the rat io as cons t an t a l lows a s i ng l e phase chang e t i me pred i c t ion
me thod for one of the basic shapes to be equa l l y appl i cabl e to the
other two , wh ich i s a des i r ab l e feature of any predict ion me thod .

Data ar e a va i lable for fr ee z i ng of foods wher e l atent heat i s r e l eased


over a range of t emperatures , therma l propert ies change with
t emperature and sens i b l e h eat effects b o th above and below the fre e z i ng
temperature rang e are s ign i f icant ( Cl e land & Earle 1 97 7 a , 1 97 9a ) .
Pred i c t io n of Thaw i ng T imes 1 49

W i th i n the tol er ance of the data , previous s tudies ( C leland & Ear l e
1 98 2 b , Pham 1 98 4 a ) sugges ted that there was n o ev idenc e that the rat io
was not constant at 6 : 3 : 2 . The pred i c t ions for al l the shapes showed
no trends w i th changes in shape or free z i ng cond i t ions ( Pham 1 98 4c ) .

The hyp o thes i s that the r a t io o f t slab = t cyl = t sph i s cons tant at 6 : 3 : 2
for b o th freez i ng and tha w i ng o f foods and that i t i s indepe nden t o f
env ironmental cond i t ions was tested mo re r i gorou sly b y use of n ume r i ca l
me thod s . The Tylose freez i ng data ( Cleland & Ear l e 1 97 6 b , 1 977 a ,
1 97 9 a , 1 97 9 b ) and the Ty lose thaw ing da t a ( Chap . 6 ) for th e slab ,
inf in i te cy l i nder and spher e shapes were pred i c ted us i ng the three
numer ical methods d i sc ussed in Chap . 4 - the f i n i t e d i fference method
and the two f i n i t e el ement me tho d formulat ions . For each exper ime ntal
run n i ne pred ict ions were made one for each comb i na t ion of the thr ee
numer ical me thods and the three geometry descr i p t ions . Each pred i c t ion
was mod i f ied by the ap prop r i at e rat i o and th e perc entage d i f ference
from the exper i mental t i me calculate d . For exampl e , an inf in i te
cy l i nder exper imen t al run was pred icted by results from slab , inf in i t e
cyl inder and spher e vers ions o f the three numer ical methods mu l t i p l i ed
by 0 . 5 , 1 . 0 and 1 . 5 respec t i vely . The perc ent age d if ferences were
ca lculated from :
pred ic ted t i me - exper i mental t ime 1 00
p ercentage d i fference (7 . 1 )
exper i mental t i me

The r es u l t s for b o th the freez i ng and thaw i ng data are summar i sed in
Tab le 7 . 1 . The f i n i te d i f ference method and the two f i n i te e lemen t
method formu l a t ions gave almost iden t ical resu lts s o only the f in i te
d i f f er enc e r esults are presented . The rat ios of n ume r i cally pre d i c t e d
t i mes :
t slab / t sph ' t slab / t cy l and t cyl / t sph w ere a lso exam i ned for
each set of exper imental runs . The mean va lues and 9 5 % con f i dence
intervals were 3 . 0 2 ±0 . 20 , 2 . 0 1 ±0 . 07 and 1 . 50 ±0 . 0 5 r espec t i ve ly for
freez ing and 3 . 00 ±0 . 37 , 2 . 00 ±0 . 1 5 and 1 . 50 ±0 . 0 7 r espec t i ve ly for
thawi ng .

The mean pred i c t ion accuracy was not s i gn i f icantly affected by the
c ho i c e o f geome t ry descr i p t i on used in the nume r i ca l calcula t ions .
Henc e , on average , the r a t io of 6:3:2 holds for b oth freez i ng and
thawi ng .
Pred i c t ion of Thawi ng T i mes 1 50

Var i a t ions of the 6:3:2 r a t i o fo r ind i v i dual exper i menta l runs aro se
from two fac tors - both the overall change in enthalpy from the i n i t i a l
temperature to a spec i f ie d f inal thermodynam i c c entre temper atur e and
the mean " e ffec t i ve" th erma l conduct i v i ty d i ffer w i th shape . These are
d i rec t l y consequ en t i a l on the d i fferent d i s t r i bu t ion of vo lume w i th
respec t to d i sp lacement from the geome t r i c centre i n the thre e shap es ,
wh i c h in turn a ffec ts the shape of the temper atur e/d i sp l ac ement
pro fi l e . These factor s mean that the t s l ab : t cy l : t sp h ra t io is
depende nt on Bi , Ste and Pk .

For freez ing the dependence of the ratio on Ste and Pk is weak . For
the ranges over wh ich Bi , Ste and Pk v ary in t yp ical free z i ng
operat ions ( Tab le 6.1 ) , there is no need to take account of these
trend s . The ratio can be cons idered constant at 6:3:2 w i thout
introduc ing any s i gn i f icant add it iona l unc ertainty into the
pred i c t ions .

For thaw ing the e f fec t o f Pk is ins ign i f icant because va lues of Pk
typ ically encoun t e r ed are sma l l ( Tabl e 6.1 ). Also thermal conduc t i v i ty
i s h igh and tempera ture d r i v ing forces are large when the sens ible heat
r epresented by Pk is transferred so th i s h eat transfer i s eas i l y
ach ieved. Converse ly , th e sens ible heat represented by Ste is
transferred mor e slowl y as temp erature d i f ferenc es a n d the thermal
conduc t i v i t i es are both lower at th is st age in the thawing process .
Typ i ca l ly i n thaw i ng Ste v a r i es over a wi de range so the Ste e f f ec t i s
larger . The B i e ffec t r e s u l t s from changes in the therma l conduc t i v i ty
wi th temper atur e and v ar i a t ion of the surface temperature and
temperature/d isp lacemen t p rof i l e at the end of the phase change p rocess
as 8i changes .

Al though d i sc ernible i n the numer ical resu l t s , the Bi , Ste a nd Pk


effects were suff ic iently smal l compared wi th the exper i mental
uncer ta inty in the data se t ( Tab les 6.2, 6.3 and 6.4) that they could
be i gnore d in the pr e sent wor k . By u se o f the 6:3:2 r a t io a l l the
slab , i n f i n i t e c y l i nder and sphere data coul d be grouped t og e ther for
analys i s by a un i f ied approach . If mor e accurate thaw i ng d ata were
ava i l ab l e , parame t ers r e la t i ng 8 i , Ste and Pk to shape may be needed t o
g i ve accurate pred i c t ions for a l l three shap es .
Pr edict ion of Thawing T i mes 1 51

1.2 PREDIC T I O N BY NUM ERICAL M ETHODS

A w i de r ange of numer ical so lut ions to thaw i ng prob lems were d i scu ssed
in Chap s . 2 and 4 . The thr ee t i me l ev e l Lees ' f i n i t e d i fferenc e scheme
and f i n i te e leme nt method that ac counted for therma l prope r t i es
con t i nuou sly var i ab l e wi th t emper ature , were shown to be the bes t
numer ical me thods for pred i c t ion of phase change i n foods .

U s i ng the therma l da t a g i ven i n Tab l e 5 . 1 the pred ict ions by the f i n i t e


d i fference me thod and the two f in i t e e l emen t formul a t ions a r e compared
w i th the ex per imen t al slab , i n f in i te cyl inder and sphere thaw i ng t i mes
i n Tab l e 7 . 3 . Ty p i ca l pred i c ted temperature pro f i l es ar e shown in
F i g . 6 . 1 to 6.3. I n a l l cases an evenly spaced 1 1 node gr i d ( w i th 1 0
two nod e l i nea r e l ements for the finite e l eme nt method s ) was u sed .
Numer i c a l ap prox i mat ion error cou ld be reduced by using a more re f ined
grid ( and/or h i gher order e l ements for the f i ni te e l eme nt method ) .
Test ing of d i f ferent grid s i z es was car r i ed ou t for some runs but as
s i gn if ican t ly d i fferen t pred i c t ions d i d not resu l t , the 11 no de grid
was cons i de red sa t i s factory .

The 95% conf idence l i m i ts of the percentage d i fferences ( Eq . ( 7 . 1 ) )


be tween the calculated thaw i ng t i mes and the exper ime ntal r esu l ts for
a l l three shap es are - 9 . 3 % to 6 . 7 % for the f i n i te d i f ferenc e method ,
- 9 . 6 % to 7 . 9% for the ful l f in i te e l ement formu l a t ion ( Eqs . ( 4 . 2 ) to
( 4 . 5 ) ) and -9 . 5 % to 9 . 3 % for the s i mp l i f ied f i n i te e l ement formulat ion
( Eq s . ( 4 . 5 ) to ( 4 .8) ) . These unc erta i nty l im i ts i nc l ude con t r ib u t i ons
from thr ee sourc es - exper imental error , therma l data error and errors
in the app l i c a t ion o f the numer i ca l method s .

Apply ing the cr i ter i a of C l e land & Earle ( 1 984 a ) t o compari sons of the
numer ical pred i c t ions wi th the exper i mental resu l t s showed no trends i n
the data that m i ght suggest maj or systema t ic exper imental o r thermal
data errors . The exper i mental error bounds cou l d no t be determined
accurately but were es t i mated to be between ±5% and ±1 0 % ( Chap . 5 ,
Tab l e 1 2 . 2 ) for the slab , inf i n i t e cyl inder and sphere thawi ng
exper i me nt s . Th i s is of the same order of ma gn i tude as the total
l imi t s for the nume r i c a l pred i c t ions , sug ges t ing that only m i n imal
uncert a i n t y has ar i se n in the ap p l i ca t ion of the nume r i c a l me thods .
Pred i c t i on of Tha w i ng T i mes 1 52

Th is was ex pec ted because relat i vely fine space and t i me g r i ds cou l d be
u sed ( Chap . 4 ) .

Some slab runs had a cont r i bu t ion from edge heat transfer tha t was
es t imate d to exceed 1% ( Sec . 5 . 4 . 4 , App . B ) . The pred i c ted thaw i ng
t imes calculated by the thr ee n umer i ca l methods for these runs when
compared w i th the expe r i mental data agreed to w i th in 1 . 3 % ±8 . 4 % , 0 . 2%
±8 . 8% and 0 . 2% ±8 . 8 % respec t i vely . If edge heat tr ansfer was
s i gn i f icant compared wi th sourc es of random error then the means of
these percent age d i fferences shou ld be not i cab l y off set from the mean
of a l l the other p ercentage d i ff erenc es . I t was conc luded that the
e x tra exper imenta l uncer t a inty in these runs had no t s i gn i f i cantly
i ncreased the overall unc erta i nt y .

Numer ical me thods , i f formulated and impl emen ted correc t ly , are the
c lo s es t to exac t pred i c t ion me thods . Ther e fore as there is no
s i gn i f icant and iden t i f iab l e systemat ic tr end or correlat ion in the
expe r i mental data the error bounds of the nume r i c a l pred i c t ions a re the
best es t imate of the ov erall random ex p e ri menta l error ( due to
uncertainty in control and measuremen t of exper i mental cond i t i ons ) .
Pr e d i c t ions be tter than thos e achieved by the nume r i ca l methods cannot
be expec ted if the same exper i mental and therma l data ar e use d unl e ss a
more exact or soph i s t icated form of the nume r i c a l methods is used
( Heldman 1 983 , Cle l and & Ear l e 1 98 4a ) . Compa r i ng the numer i ca l
pred i c t ions for each shape i n d i v idua l ly ( Tab l e 7 . 3 ) g i ves est i ma t es of
the expe r i mental error for the exper i ments w i th each shape . The
average 95% confi dence bou nds of the observ ed d i fferences for the thr ee
nume r i c a l methods are on average ±7 . 1 % , ±5 . 0 % and ±1 1 . 3 % for slab s ,
inf i n i te c y l i nder s and s pheres respec t i ve l y . As expected the
uncer t ainty for s l ab s and i n f in i te c y l inders i s lower than for s pheres
as b e t t er control and measuremen ts of the exper i mental cond i t i ons ,
espec ially d imens ions and surface heat transfer coe f f i c ients , were
pos s ib l e ( Chap . 5 ) . The above uncerta i nt y b ounds are co nsi sten t w i th
exper i mental variab i l i ty measured by summing that measured i n rep l i cate
exper i men t s , and systema t i c error from other sources not ev i dent from
the rep l i c ates ( for examp l e , measurement of h) .

The mea n p r ed i c t io n error a r i s i ng from cal c u l a t ions by the s i mp l i f i ed


Pred i c t ion of Thaw ing T i me s 1 53

f in i te e lement me thod formu l a t ion was h igher than for the oth er two
methods . Th i s is due to the relat i vely crude way in wh ich variat ions
in the therma l proper t i es are i ncorpor at ed in to th i s method . For th is
method increas i ng the number of nodes and e l ements d id have some
benef i ts and reduced the d i f ferenc e in pred i c t ion compared wi th the
other me thods . I t is debat ab le wh ether the sav ing s in c omputat ion
cos ts j us t i f ies th i s loss in ac curac y as the decr ease in compu ter
memory s i z e requ iremen ts and increase in speed achie ved by us ing th e
s i mp l e formu l a t i on fo r one - d ime ns iona l work i s sma l l .

Centre temperature prof i l es were accurat e ly pred icted by all th e


numer ical methods though , for the reason out l ined ab ove , th e s i mp l i f i ed
f i n i t e e l ement formulat ions per formed l east well in th is respec t .
D i f ferenc es be tween pred icted and exper imental t emperatures were more
l i k ely to have ar i sen from uncertainty in thermocou p l e pl acement rather
than nume r i c a l me thod error . In Chap . 5 i t was shown that th ermo coup l e
placement was a major problem . It d id not affect thaw ing t i me
pred i c t ion ( because the breakpo in t ana ly s i s was used ) , but did g i ve
s i gn i f icantly d i fferent exper imenta l temperature prof i l es from those
expect ed at the c entr e .

A l l three nume r i c a l me thods t ended to predict sur face temperatures


cons i s tent l y lower than tho se measured expe r i mental ly for the i n f i n i te
cy l i nder and sphere expe r i ments . Th is was due to the prob lems w i th
therma l contac t , a ir v o i ds and thermocou p l e p lac ement at the surface
for these shapes . As d i scussed i n Secs . 5 . 5 and 5 . 6 the methods use d
t o es t i mate the average surface heat transfer coeff ic i e nt onl y u sed the
centr e temperature and not the sur face temperatur e dat a . The
exper i ment a l l y de term i ned surface t empe ratur e was t aken from the
thermo cou p l e that exh ib i t ed the mos t rap id change in t emperatur e . Th i s
the rmo cou p l e was not nec es sari l y r epresentat i ve o f the t ru e av erage
surface t emperatur e . Numer ical pre d ict ions made w i th the average
surface heat transfer coe f f i c ient therefore under es t i mated the rate o f
change of t h e expe r i mental values . For the slab shape the surfac e
temperatur e pred i c t ions were cons i st en t l y more accurate as problems
w i th voids were a vo i de d .

For the f i n i t e e lement methods some unex pect e d d e v iat ions of the
Pred i c t i o n of Thawing T i mes 1 54

temperature prof i l es were ob served when e l emen t s wer e near th e l at ent


heat temperature rang e . Th is occurred because of the way the finite
e lemen t methods incorpor ated the therma l proper t i es and was an e f f ec t
of the int egral nature of the f i n i te e l ement method , espec ia l ly wh en
therma l property change w i th temperatur e are rap i d ( Cle land et al
1 98 4 ) .

7.3 PRED ICTION BY SIMPLE FORMULAE

7.3. 1 E x i s t ing Pr e d i c t ion Formulae

In Chap . 2 a large number of s i mple formulae for calculat io n of


freez i ng and thawing t i mes i n slab s , inf ini te cyl inders and /or spheres
were d i scusse d . Many of these formulae were not ex pec ted to be
accurate i n . the s i tuat ions cons i dered in th i s work . So l u t ions that
only take account of the f irst k ind of boundary cond it ion are not
sui tab l e for genera l use as they g i ve poor pred i c t ions excep t where the
surface heat trans fer coe f fic ient is extreme ly h igh . These types of
methods were there for e no t cons i dered any fur ther . S i m i l ar l y so l u t i ons
for the second k ind of boundary cond i t ion wer e cons ider ed imprac t i cal .

Thaw i ng t i mes for the slab , i nf i n i t e cyl inder and sphere expe r i me nta l
data were calculated u s ing the me thods that cons i dered the th ird k ind
o f b oundary cond i t ion and that seemed most l ik e l y to l ead to r easonab l e
pred i c t ion accuracy . The percent age d i fferenc es between th e calculated
r esu l ts and the exper i mental thaw i ng t imes were found (Eq. (7 . 1 ) ) . A
summary of these values i s g i ven i n Tab le 7 . 2 .

Mos t of the me thods t es t ed assume that al l th e l aten t heat is ab sor b ed


at a un i que thawi ng temperature and that the froz en and unfrozen phase
have constant therma l p roper t i es ( thermal conduct i v i t y and s p ec i f i c
heat capac i ty ) . For these methods the thermal conduct i v i ty o f the
comp letely un froze n mat erial was used , and the l at ent heat comp onent
was ob ta ined by sub tract i ng the sens ible heat component from the total
enthalpy change between 0 °C and - 1 0 °C . Excep t where a " mean" thawing
temperatur e was def i ne d , the un i qu e thaw i ng temper ature was t ak en as
the i n i tial freez i ng temperature because th i s is wel l de f ined for mo s t
foods and the r a t e of change o f the i ce frac t io n w i th temperature i s a t
Pred ict ion o f Thawi ng T i me s 1 55

a max imium at th i s temperatur e . The cho ices o f proper t i e s made f i t the


concep t of the analy t ical me thods , are no t amb iguous , and a lso use the
b es t k nown the rmal data for most foods . The thermal p roperty v a l ues
used are g i ven i n Tab le 5 . 2 .

Some methods based on mod i f icat ions to e x i s t ing analy t i c a l formulae


de f i ne the therma l pr oper t i es d i f ferent ly , general l y to mor e c lose l y
approx ima t e the true therma l proper t i es dur i ng the phase change
process . Common ly , the enthalpy change is de f ined to inc lude sensible
heat e ffec t s , and average therma l conduct i v i t i es and mean phase change
temper atur es are used . In each such case the gu i de l ines fo r therma l
proper ty determina t ion sugges ted by the proposer of the pre d i ction
me thod were fo l lowed as c losely as poss ibl e .

Some emp i r ical formulae developed spec if ically for freez ing t ime
pred i c t ion wer e a lso tested . If it was poss i b l e the ana l ogous
emp ir ical formula was deve loped for thaw ing . Generally the methods
that could be adap t ed wer e those wi th a more comple te theor e t i ca l
bas i s . Other formulae were s i tuat ion o r p roduc t spec i f ic and cou ld not
be made app l i cab l e .

Pred ict ion methods that d id not pre d i c t the exper i mental thaw i ng data
wel l wer e l im i ted i n one or mor e of the fol low i ng way s .
( 1 ) Methods developed for semi - i n f in i te slab s cannot be app l ied to
f in i t e shape s if sens ibl e heat e f f ec ts p r io r to phase change are
s i gn i f icant . Th is is because heat transfer is ca lculated to occur
from a greater volume i n the semi - i n f i n i te s l ab than i s ac tua lly
present in the fi n i te geometry ( Geuze e t a l 1 97 2 , C l e land 1 97 7 ) .
( 2 ) Me thods assuming a un ique thaw i ng temper ature ( t ak en as equ a l to
the i n i t ial freez ing t emperature ) lead to overpre d ic t ion of thawing
t i me because much of the latent h eat ( as we l l as subcoo l i ng
sens i b l e heat ) i s ab sorbed over a range o f temperatures l ower than
Tif • There for e the ac tual temperatur e d r i v i ng force for h eat
transfer is greater than tha t used i n calculat ions .
( 3 ) Me thods that used the unfrozen phase thermal conduc t i v i ty dur i ng
the phase change p rocess t ended to overpred i c t the thawing t i me s .
The t ru e mean therma l conduc t iv i ty dur i ng the tha w i ng process is
greater than or equal t o the unfrozen value s o mor e heat t ransfer
P r ed ic t i on of Thaw i ng T i mes 1 56

occurs in prac t i c e than calculated us ing the fully un frozen va lue .


( 4 ) Thaw i ng t i me pred i c t ion me thods ignor i ng sens ibl e heat added to the
thawed reg ion tend to underpre d i c t .
( 5 ) Me thods ignor i ng the sensible heat added to the frozen reg i on
underpre d ict . The extent of the underpredict i on from th i s cau se i s
l ess than that from ( 4 ) .

Numer ical me thods that so lve the govern ing part i a l d i f feren t i a l
equat ion for heat transfer and t a k e accoun t of temperatur e dependent
therma l prop e r t i e s correc t l y model the thawing process phy s i ca l ly and
there for e g i ve pred i c t ions that are not affec t ed by the above f ac tors .
Compar i son wi th the results for the nume r i c a l me thods he lps to i dent i fy
how wel l a pred i c t ion method performs in try i ng to g i ve a phy s i c a l l y
correc t descr i p t ion of these factor s and how they af fect thawing t i me .
I t i s conven ient to u se the corre l a t ion coe f f i c ient (r) compar i ng
percentage d i fferences in the manner of C l e land & Earle ( 1 98 4 a ) .

The methods and resu l t s in Tab l e 7 . 2 were d iv ided i nto n i ne group s .


The Group A methods are a ffec ted b y the f ir st o f the abov e l i mi tat ions
and ar e no t accurate . O ther me thods that so l ve for the sem i - i n f i n i te
slab do no t account for in i t ia l subcoo l i ng and cons equen t l y predict
d i f feren t ly to Gr oup A methods .

Plank ' s ( 1 9 1 3 ) equa t ion ( Group B) was l im i ted by the l a tter four
factors . The compe nsatory natur e of ( 2 ) and ( 3 ) aga ins t ( 4 ) and ( 5 )
means that the mean p re d i c t ion error was + 6 . 0 % . Howe ver , the spread o f
the p r e d i c t ions was l arge ( ±4 3 % at the 9 5 % l ev e l o f conf i denc e ) and the
correlat ion wi th the f i n i t e d i fferenc e method r esu l t s was l ow . Th i s
i nd icates that the compensa t i on was not cons istent between runs w i th
d i fferent cond i t ions , render i ng the equa t ion unre l i abl e . In free z i ng ,
Plank ' s equa t ion cons i s tently underpredicted by 20% to 40% b ecause the
equ i valent prob l ems to f actors ( 2 ) to ( 5 ) a l l l ead to underp r ed i c t i on
for that process .

The analyt i c a l methods due to Rutov ( 1 936 ) , Goodman ( 1 958 ) , Sh ih & Chou
( 1 97 1 ) , Sh i h & T say ( 1 97 1 ) , Huang & Sh i h ( 1 975a , 1 975b ) , Ker n ( 1 977 ) ,
Glas s er & Kern ( 1 97 8 ) , Yan & Huang ( 1 979 ) , Cly ne & Garc ia ( 1 9 80 ) , Cho &
Sunder l an d ( 1 98 1 ) , Sol i man ( 1 98 1 ) and Hi l l & Kuc era ( 1 9 83 ) , and the
Pred i c t ion of Thaw i ng T i mes 1 57

emp i r i ca l methods of Bax ter ( 1 96 2 ) , Tao ( 1 967 , 1 968 ) and Good l ing &

Khader ( 1 97 4 ) , form Group C . They ar e a l l l i m i ted in s imi lar way s to'


Plank ' s equat ion excep t tha t they t ak e ac coun t of the sens i b l e heat in
the unfrozen phase . Al l methods w i t h in th is group gave s i m i lar
p red i c t ion accuracy . The results fo r Goodman ( 1 958 ) shown in Tab l e 7 . 2
are t y p i c a l o f the whole group . The mean d i f ferenc e was h i gher than
that for P lank ' s equa t io n and the spread i s red uced because factor ( 4 )
was t ak en into account . The predict ions are st i l l poor b ecause of the
o th er l i m i t at ions .

Group D ar e the mod i f ied ana l y t ical methods ( semi-analy t i ca l ) tha t are
based on mu l t ipl icat i ve factors to ac coun t for sen s i b l e heat e f f ec ts
( factors ( 4 ) and ( 5 ) ) . They all tend to sub stan t i a l ly ov erpre d ic t
thaw i ng t i mes , and pred i c t ions tend to ha ve a la rge spr ead .

Group E methods u se e i ther a mean phase change tempera ture ( Mo t t 1 96 4 ,


Flem i ng 1 967 ) , an avera ge thermal conduc t i v i t y ( Mel lor & Sepp i ng s 1 976 )
or both ( Mod i f ie d Plank ) to take account of fac tors ( 2 ) and ( 3 ) . Such
me thods s t i l l do not t ak e accoun t of a ll the l i m i t a t ions in a
phy sically rea l i s t i c manner so the spread of p re d i c t ions was s t i l l
l arge .

The Gro up F methods break the phase change p ro cess into three s tages
and at temp t to appro x i mate the heat transfer in each s t age . They
per form b e t ter than Group D because they mod e l the true phy s ic a l
cond i t ions mo r e rea l i s t i cal ly . To rema in s i mp l e these methods use some
avera ging t echn i ques . They tend to overpre d i c t th e thawing t i me s
usual l y b ecau se they do no t t ake accoun t of factor s (2 ) an d (3 )
proper ly .

The emp ir ic a l formulae deve loped for freez ing ( Group G) all tend to
overpred i c t thaw i ng t i mes as the emp i r i ca l corr ec t ion factor s are not
approp r i a t e for thaw i ng . The method of Succar & Hayakawa ( 1 984 ) was
not t es te d as it is poss i b l y unnec es sar i l y complex and y et was not
expected to be mor e accurat e , than the other me thods in th i s group .

The two emp i r i cal methods ( Gr oup H ) spec i f i cal l y developed for thawi ng
t i me pred i c t ion gave the bes t pre d i c t i o n ac curacy . Creed & Jame s
Pr ed ic t ion of Thaw i ng T i mes 1 58

( 1 98 1 ) gave an emp ir i c a l formul a for thawi ng o f slab s o f bone less bee f


based on b oth numer ical pred i c t ions and expe r i menta l thaw i ng dat a . It
gave pre d i c t ion w i th 95% conf idence bounds o f -1 7 .7% to 18.1%. The
me thod i s produc t s p ec i f i c because it does not t ake ac c ount of
d i fferent thermal proper t ies ( k 1 , k s , c 1 , C s , � . T i f ) . A l s o the
pred icted thaw i ng t i me is only va l i d for T i n =-30 °C . Ty lose has th erma l
proper t i es s im i lar to those of b ee f a nd thaw i ng t i me i s on ly weak l y
dependent o n changes in i n i t i a l temperatur e . Therefore the predict ions
have a mean d i ff erenc e of z e ro but a s i gn i f i can t l y gr eater var i ab i l i t y
than wou ld o ccur i f these factor s were t ak en into ac count .

Ca lvelo ( 1 9 8 1 ) p r esen ts an emp ir ical formu l a based on Pl ank ' s equ a t i on


der i ved by regress ion ana l y s i s of n umer ical predict ions for thawi ng o f
bee f s l abs . Though k 1 and c 1 ar e incorpor ated in th e equ a t i on , the
other p rope r t i e s ( �H . T i f ' k s , C s ) do not need to b e spec i f i ed and the
method is therefore product spec i f ic . As Ty los e and beef ha ve s i m i lar
the rma l proper t i es th is method gave accur a te pred i c t ions w i th error
bounds of - 1 1 . 8 % to 1 1 .8% at the 95% l ev e l of conf i de nce . The
corre l a t i on of the p ercentage d i fferences wi th the f i n i te d i f f erence
pre d i c t ions was h i gh , suggest ing that a l l maj or sourc es o f thaw i ng t i me
v ar i a t i on ar e accoun t e d fo r by th i s formu l a . The res i dual pred i c t i on
uncer t ainty , apart from that due to ex per i men tal error , was sma l l in
magni tude and was not s i gn i f ican t l y correlated t o any other ind i v idual
fac tor .

W i th in Group I , on ly th e me thod due to Golov k in et al ( 1 974 ) gave


r easonabl e p r ed i c t i on accuracy . As the pred i c t ions were no bet t er than
those for some other s i mp l er me thods , and because the publ i shed
der ivat ion i s unc l ea r , the method was not cons i dered fur ther .

7.3.2 Improved Pre d i c t ion Methods

Of the ex i s t i ng methods t ested , only three ( Creed & Jame s 1 98 1 , Calvelo


1 98 1 and Pham 1 98 4 a ) gave p r ed i c t ions of r easonab l e accuracy and
re l i ab i l i ty ( mean percent age d i f ference c lo s e to zero a nd a low
standard dev i a t ion of the percent a ge d i f ferences ) . Fur ther , they took
account of most of the sources of var i at ion in thawi ng t i me s and wer e
consequent l y h i gh l y corr el ated wi th resu l ts from nume r i c a l methods .
Pred i c t ion of Thaw ing T i me s 1 59

The b es t methods were those due to Calvelo ( 1 98 1 ) and Creed & James
( 1 9 8 1 ) but both have the d isadvantage that they do not t ak e acc o un t of
d i f fere nt therma l proper t i es for d i fferen t foods tuf fs . Creed & James '
method was also indep endent of i n i t ial temperature . Pham ' s ( 1 9 8 4a )
me thod has the advant age of hav i ng a pre dom inate l y ana ly t ical bas i s .
I t has a low standard deviat ion due to the three stage approach
cons i der i ng all sources of thawing t i me v a r i a t ion but the mean
perc entage d i fference was offset from z ero.

Becau se a l l the e x i s t i ng methods have weaknesses i t seemed worthwh i l e


t o seek a new pre d i c t ion me thod . Emp ir i ca l ly mod i f i ed forms of Pl ank ' s
equat ion s i m i l ar to those found to be u se ful fo r freez i ng ( Cleland &

Ear le 1 98 2 b , Hung & Thomp son 1 9 83 ) were invest igated because of the ir
s i mpl i c i t y and succ ess for freez ing t i me p r ed i c t ion . Also p r ed i c t i on
formulae s i m i lar to those of Calvelo ( 1 9 8 1 ) and Pham ( 1 984a, 1 9 84c )
were inves t i gated to see i f they could be impr oved and made appl i cable
to a w i der range of s i tuat ions . The Creed & James ' ap proach may have
y ie l ded an accurate product - spec i f ic formul a bu t was not cons idered as
su itable for de ve lopment of a general formula compared w i th th e other
me thods .

The fo llow i ng four predict ion formulae were found to be super ior to
o thers t r ied . The i r pred i c t ions are summa r i se d i n Table 7 . 3 .
( a ) Use o f the bas i c form of Calv e lo ' s ( 1 9 8 1 ) solution b ut represent ing
i t in terms of Fo , Bi , Ste and Pk so that all the therma l
proper t ies a r e i nc luded . Analy s i s of the exper imental dat a by
mu l t ip l e non-l i ne ar r egr ess ion gave the fo llow i ng pred i c t i on
equa t ion :
AD
Fo - = 1 . 4 29 1
[ 0 .5 0 . 1 25
+ --
] 1 . 0 2 48
Ste
0 . 27 1 2
Pk
0 .061 0
(7.2)
2V Bi Ste Ste
( b ) Use of we igh ted mul t i ple l i near regress ion to f ind mod i f ie d
formul ae for the shape factors i n Plank ' s equ a t ion i n a s i m i l ar
manner to that u se d. by C l e land & Earle ( 1 9 7 6 b , 1 97 7 a , 1 9 7 9 a , 1 97 9b ,
1 9 82b ) . The best formu l a deve l oped was :
AD p R
Fo - • + -- (7.3)
2V Bi Ste Ste
where p - 0 . 5 ( 0 .7754 + 2 . 2828 S t e Pk )
R - 0 . 1 25 ( 0 . 427 1 + 2 . 1 220 S t e - 1 . 48 4 7 Ste 2 )
Pred i c t i on of Thawi ng T i mes 1 60

Other t erms were stat i s t i ca l l y s i gn i f i cant in thi s model but did


not l ead to wor thwh i le reduc t i on o f the standard de v i at ion o f the
percentage d i f ferences between exper i ment a l and pred i cted thawi ng
t imes . These terms may have been f i t t i ng a systema t i c component of
the exper imental er ror .
( c ) A thr ee stage ap proach s i mi lar to Pham ( 1 98 4 a , 1 98 4c ) . The best

[
pred i c t i on formula of th i s type was :

J
AD 3
Mi i D hD
t
2V
I T 2h
1
--
1 + -
4k i
(7. 4 )
� i
whe re llH 1 C s ( T i fave - T in )
�T 1 T a - ( T in + T i fave > 1 2
k1 ks
�H 2 L
�T 2 T a - T i fav e
k2 0 . 25 k s + 0 . 75 k 1 = 0 . 83 for Tyl ose
�H 3 C l ( T ave - T i fave )
�T 3 T a - ( T ave + T ifave > 1 2
k3 kl
T ave average f i na l temperatur e
= T r i n - ( T f i n - T a ) / ( 2 + 4k l /h/ D )
T i fav e = T i f - 1 . 5 = - 2 . 1 for Tyl ose
Pred i c t ion accuracy was sens i t i ve to the values of the mean
freez i ng temperature , the average thermal conduc t i v i ty i n the phase
change per iod and the average f i na l temperature . To retain the
ana l y t i cal bas i s the val ues of the mean thaw i ng temperature and the
avera ge f i nal temperature were calculated by the methods suggested
by Pham ( 1 98 4 a ) . For accurate pred i c t ions i t was nec essary to
emp i r i ca l l y f i t the value for the mean thermal conduc t i v i t y in the
phase change p er iod The we i ghted val ue for k 2 of 0 . 83
w m - 1 oc - 1 for the phase change per i od , although phy s i cally
reasonab l e , has no bas i s apar t from conven i enc e . Equat ion ( 7 . 4 ) is
c lose ly related to the methods of Pham ( 1 9 8 4a , 1 9 8 4c ) , wh i ch are
both of s im i la r pred i ct ion accuracy for freez i ng . Equat ion (7.4)
was cons i dered to be the emp i r i ca l thawi ng equ i valent of the Pham
( 1 9 8 4 c ) freez i ng approach , rather than the equ i valent of the Pham
( 1 98 4 a ) method as the l at ter has mor e substant ial ana ly t i cal bas i s .
( d ) D irect f i t t ing o f a correct ion t o Plank ' s equat ion gave :
Pred i ct i on of Thaw i ng T i mes 1 61

AD [ 0.5 0 . 1 25 1 �H
*
[ 0 .0244 Pk
+ 0 . 6 1 92 -
j ( 7 .5 )
Ste j
Fo = + 0 . 89 4 1 -
2V B i Ste �H Ste Bi
* 3
where �H =entha l p y change f r om - 1 0 °C to T (J m- )
ave
Equa t ion ( 7 . 5 ) i s s i m i l ar i n form to the free z i ng t ime p r ed i c t i o n

method pro posed by Pham ( 1 9 83 ) . The Pham pred i c t i o n formula has

[ ]
not been publ i sh ed i n the l i te r a t ur e . I t is :

AD 0.5 0 . 1 25
Fo - = + -- [ 1 + 0 . 0 4 /( B i Ste ) + 1 . 2 7 /( Pk St e ) ] ( 7 . 6 )
2V Bi Ste Ste

A l l four poss i b i l i t i es ga v e accurate p r ed i c t i ons and th e p re d i c t i o ns

take account of all maj o r sourc es of va r i at io n in thaw i ng t i me as

i nd i ca t e d b y the h i gh c o r r e l at ion w i th the f i n i te d i ffer enc e method .

Al l the me thods are s i m p l e and have some phy s i ca l bas i s th ough the i r

der i v at ions w e r e no t analy t i cal .

Further r e f i neme n t o f these methods may be poss i b l e . It is unl i k e l:(

that s i gn i f i cant i n crease in p r ed i c t ion a ccuracy cou l d b e ach i eved

w i thou t i n t roduc i ng fur the r terms ( tha t could be f i t t i ng s y s t ema t i c

exper i mental error ) , or emp i r i ca l l y f i t t i ng the e q uat ions s epar a t e l y

f o r e a c h of the three sha pes . F o r a l l four formulae abov e , the s p r ead

o f pr ed i c t ions for each of the three shapes was s l i gh t l y h i gher than

for the n ume r i ca l methods . Th i s i s because the u n i f i ed app roach us i ng

the 6 : 3 : 2 r a t i o i s no t compl e t e l y accurat e . I f data for each shape are

ana l y se d separat e l y , thr ee sl i ght l y d i f f e r en t forms of Eqs . (7.2) to

( 7 . 5 ) , s p ec i f i c for e ach shap e , can be der ived . The p e rc entage

d i f f e r enc es for these sha p e s pec i f i c formulae have s i m i lar s t andard

dev i a t ions to the nume r i ca l me thods for each shap e .

The pos s i b i l i t y must s t i l l e x i s t that a d i f ferent ap proach to any of

tho s e i nv es t i ga t ed cou l d conc e i v ab l y l ea d t o a mo r e a c c ur a te formu l a ,

but th i s i s not cons i de r ed l i k e l y .

Because Eqs . ( 7 . 2 ) to ( 7 . 5 ) were d er i ved for T y l o s e thaw i ng d a t a on l y ,

they are r est r i c ted i n prat i c e to tho se foods wi th therma l proper t i es

s i m i l ar to those of T yl o s e . Th i s i n c l udes a w i de range of med i um to

h i gh mo i st ure foods and is not unduly l i mi t i ng . O ther r anges of

app l i cab i l i ty are :

0.6 < Bi < 57 . 3


Pred i c t ion of Thawi ng T i me s 1 62

0 . 0 85 < St e < 0 . 7 68
0 . 065 < Pk < 0 . 27 2

7 .3.3 Compa r i s o n W i th Freez i ng T i me Pr ed i c t ion Formu l ae

Me thods s i mi l ar in p r i nc i p l e to Eqs . ( 7 . 3 ) to ( 7 .5 ) have been


successfu l ly used for fr eez i ng t ime pred ict ion ( Cl e land & Ear l e 1 98 2 b ,
?ham 1 9 83 ( Eq . ( 7 . 6 ) ) , ?ham 1 9 84 c ) . The ma in advan t ages of th ese t ypes
of me thod ar e the ir s i mp l i c i ty , acc uracy and min ima l requ ir ement for
therma l property data .

Zar i tzky et al ( 1 982 ) use d an ap proach for freez ing s im i lar to


Eq. ( 7 .2 ) , bas i ng it on numer ical ly pred i c ted d ata . The method
pred icted the da t a of C l e land ( 1 977 ) poor ly so mult iple no n-l i near
r egr ess ion was used to der i v e a mor e accurate formu l a of th i s ty pe for
freez ing :

Fo
AD
- = 1 .3179
[ 0 .5 0 . 1 25 0 . 957 6
+ -- Ste
]
0 . 0 5 50 0 . 00 1 7 Bi +0 . 1 7 2 7 Pk
10 (7 .7 )
2V B i Ste Ste
Equat ion (7 . 7 ) has a mean pre d i c t ion error of 0 . 1 % ±7 . 3 % at the 95%
conf i dence level and a corre lat ion coe f f i c ient o f 0 . 61 compared w i th
the f i n i t e d i f ferenc e method pre d i c t ions errors , when tested aga i nst
the freez i ng data of Cle land & Ear l e ( 1 977 a , 1 979a , 1 9 79 b ) . Th i s
compares favourab ly wi th th e pred ict ion acc uracy of the other prov en
freez i ng t i me p red i c t ion formulae ( C l e l and & Ear l e 1 9 82 b , Pham 1 9 8 3 ,
1 98 4 a , 1 98 4c ) .

C l e land & Ear l e ( 1 9 8 4a , 1 98 4 b ) and Pham ( 1 983 , 1 98 4 a , 1 9 8 4c ) showed


that the ir me thods were superior to other rec e nt fr eez i ng t ime
pred i c t ion me thods due to Mascheron i & Ca lve lo ( 1 982 ) , de M i che l i s &

Ca lvelo ( 1 983 ) and Hung & Thompson ( 1 9 83 ) for a l arge comp os i te


freez ing data se t . Cons equent ly , the number of accurate simple
freez i ng t i me p r ed ic t ion me thods for fur ther s tudy was reduced to four .
These were that of C l e land & Ear l e ( 1 982 b , 1 98 4a ) , tha t of ?ham ( 1 98 3 )
( Eq . ( 7 . 6 ) ) , that of ?ham ( 1 98 4 c ) and Eq . ( 7 . 7 ) . The method of ?ham
( 1 98 4 a ) was not cons i dered further as i t performs no bet ter than the
?ham ( 1 984c ) method . S im i lar ly Eqs . ( 7 . 2 ) to ( 7 .5 ) , wh i ch are the
thawi ng equ i valent of the these four freez i ng me thods , wer e the only
s i mp l e thaw i ng t i me pred i c t i on me thods cons i d ered fur ther .
Pred i c t ion of Thawing T i mes 1 63

7.4 SUMM ARY

W i th in the accuracy of the pres ent freez ing and thaw i ng data , it was
val i d to t ak e the ratio of phase change t i mes fo r s l ab s , in f in i te
cyl i nders and spheres to be cons tant at 6:3:2 j ust i fy ing a un i f ied
approac h for freez i ng a nd thaw i ng t i me p red i c t ion fo r these shapes .

The three nume r i ca l methods ( the f i n i te d i fferenc e me thod and two


f i n i te e l emen t me thod formulat ions ) that ac coun ted for therma l
proper t i es cont i n uou sly var iab l e wi th temperatur e , accura tely pred icted
thaw ing t i mes for the three bas i c shapes . The nume r ical methods
pred ic ted temperature profi les dur ing thaw ing as accura tely as cou l d be
expec t ed tak ing into ac coun t the imprec i s ion of some of the measur ed
da t a .

N o p re v iously pub l i shed s i mp l e pred ict ion formu la was both suffic iently
accur ate an d su i t ab l y expr essed fo r i t to be ad op t ed as a genera l
thaw ing t i me pred ict ion me thod . Four improved formulae ( Eqs . ( 7 . 2 ) to
( 7 . 5 ) ) , that ga ve comparab l e resu l ts w i th the nu mer ical methods were
deve loped. Each of these formulae represents a d i fferent conc ep tual
approa ch i n mod i fy i ng th e wel l-known Pl ank ' s equa t ion i n way s tha t had
been previously suc c essful for fr eez ing t i me predict ion .

For both freez i ng and thaw i ng , al l the approaches y ie l ded formu lae that
are of s imi lar acc uracy and s imp l i c i ty . A l l are l i m i ted mor e by the
accuracy of the data from wh ich th ey wer e der i ved rathe r than the
inherent inaccuracy of the ap proach used . Within the l im i ts of the
t es t i ng a ga i ns t exper imental dat a , al l of these approac hes ar e of
comparab l e use fulness . It is u n l i k e ly that s i gn i f i cantly bet t er
me thods can b e deve loped that reta i n the adv antage of s i mp l ic ity .

None of the four approaches can b e recommended as be ing s ign i f i cantly


mor e accurate than the other s for the thr e e bas i c shap es . I f a method
can be eas i ly made compa t i b l e wi th t echniques to account for mor e
comp l e x geome t r y then th i s wou l d b e a u seful f eatur e . There for e , the
cho i c e between the four approaches for thawi ng t i me pre d i c t i o n i s best
de l ayed un t i l a ful l er cons ider a t ion of a r ange of geome tr i c shapes i s
made .
Pred i c t ion of Thawi ng T i mes 164

Tab l e 7 . 1 Summary of Percentage D i fferences Between E xper imental


Freez i ng and Thawi ng T i mes For S i mple Tyl ose Shapes and
Fr eez i ng and Thawing T i mes Calculated By Sl ab , Inf in i te
C y l inder and S phere Vers ions of the F i n i te D i fferenc e Method

Data 1 FDM 2 Corr 3 Mean Std Dev '+ Max '+


Type Vers ion FDM (%) (%) ( %)

s s 1 . 00 0.8 6. 1 -1 4 .6 9.3
c 0 . 98 -0 . 3 5.6 -1 4 . 1 7.4
Freez ing 5 SP 0 . 92 -0 . 8 5.2 -1 3 . 5 6.4

c s 0 . 97 -3 . 1 6.1 -1 5 . 9 5.6
c 1 . 00 -3.6 4.7 -1 4.1 3 .0
Freez ing 5 SP 0 . 93 -3 . 4 4.3 -1 3 . 3 1 .8

SP S 0.71 -2 . 6 5.4 -1 3 . 2 6.1


c 0 . 92 -2 . 8 4 .0 -1 1 8• 2.6
Free z i ng 5 SP 1 . 00 -2 . 2 3.5 -1 0 . -3 4.1

S , C , SP S 0 . 93 -1 . 3 6.1 -1 5 . 9 9.3
c 0 . 97 -2 . 0 5.1 -1 4 . 1 7 .4
Freez i ng 5 SP 0 . 93 -2 . 0 4.6 -1 3 . 5 6.5
Unmod . 1 . 00 -1 . 4 5. 4 -1 4 . 6 9.3

s s 1 . 00 -0 . 8 3.5 -6. 3 6.8


c 0.79 0.4 6.0 -8 . 5 13.o
Thaw i ng 6 SP 0 . 69 1 .3 7.9 -9 . 7 17 .7

c s 0 . 45 -2 . 4 4.1 -9 . 6 4.8
c 1 . 00 -2 . 2 2.4 -6 . 8 3 .6
Thaw i ng 6 SP 0.71 -2 . 4 3.6 -8 . 3 6 .- 1

SP S 0.64 0. 4 8. 1 -1 3 .7 1 3 .7
c 0 . 92 -0 . 2 5.9 -9 . 0 10.4
Thaw i ng 6 SP 1 . 00 -0 . 9 5.5 -9 . 5 9.1

S , C , SP S 0 . 68 -o . 9 5.7 -1 3 . 7 1 3 .7
c 0 . 86 -0 . 6 5. 1 -9 . 0 1 3 .0
Thaw i ng 6 SP 0 .78 -0 . 6 6.1 -9 . 7 1 7 .7
U nmod . 1 . 00 -1 . 3 4 .0 -9 . 5 9.1

1 S = s l ab ( 4 3 freez ing and 3 5 thaw i ng runs ) , C = i n f i n i t e c y 1 1 nd er ( 3 0


freez i ng and 3 4 thaw i ng runs ) , S P = sphere ( 30 freez i ng and 3 5
thaw i ng runs ) .
2 Version of the f in i te d i fference me thod .
' Correlat ion coeff ic ient ( r ) compared w i th the percentage d i ffer ences
for the unmod i f i e d vers ion of the f in i t e d i fferenc e me thod •

.. Std Dev a s t andard dev i a t ion , M i n • m i n i mum , Max • max i mum.


5 Data from C l e land & Ear le ( 1 97 7 a , 1 97 9 a , 1 97 9b ) .
6 Dat a from Tab l es 6 . 2 , 6 . 3 and 6 . 4 .
Pred i c t ion of Thaw i ng T imes 1 65

Tab l e 1 . 2 Summary o f Percentage D i f ferences Between Exper i mental


Thaw i ng T i mes For Ty lose Slab s , Inf i n i te Cyl inder s and
S pheres and Thaw i ng T i mes Calculated By S imp l e Pred i c t i on
Formulae
D at a from Tab l es 6 . 2 , 6 . 3 and 6 . 4 .
Total o f 1 0 4 runs - 35 for slab s , 3 4 for i n f in i t e cy l i nders
and 35 for s pheres .

Group Me thod & Re fer enc e Mean Std Dev 1 Min 1 Max 1 Corr 2
(%) (%) (%) (%) FDM

Neumann ( 1 9 1 2 ) in Carslaw & 69 . 6 11 0.5 - 68 . 8 4 1 1 . 7 -0 . 1 3


Jaeger ( 1 95 9 ) , p . 282
A
Charm & Slav in ( 1 9 6 2 ) 1 73 . o 1 42 . 6 16.5 5 3 6 . 6 -0 . 0 8
Hrycak ( 1 96 3 , 1 9 67 ) 1 68 . 9 1 47 . 0 6.9 3 4 3 . 0 - 0 . 09

B Pl ank ( 1 9 1 3 ) 6.0 21 . 4 - 28 . 0 50 . 5 0 .08

C Goodman ( 1 9 58 ) 13.3 17.6 -1 5 .3 53.7 0.1 4

Nagaoka e t a l ( 1 955 ) 79 . 1 1 5.2 45 . 4 1 22 . 6 0 . 33


Levy ( 1 9 58 ) , Edd ie & 95 . 3 21 . 7 56 . 2 1 60 . 1 0.18
Pearson ( 1 9 58 )
Ear l e & Freeman ( 1 9 6 6 ) 52 . 0 30 . 6 3 .0 1 15.2 0 . 08
D Wa lker ( 1 970 ) 38 . 2 17.9 7.9 86 . 3 0.15
Van i chsen i ( 1 97 2 ) 33 . 7 18.8 -2 . 8 78 . 1 0. 1 4
Van ichseni e t a l ( 1 972 )
Frazerhurst ( 1 97 2 ) 69 . 7 59 . 8 -1 0 . 3 30 4 . 7 - 0 . 1 6
S l a t t er & Jones ( 1 9 7 4 ) 29 . 6 1 7.0 0.3 74.0 0.14

Mot t ( 1 9 64 ) 1 2.3 1 5 .7 -1 8 . 2 49 . 7 0.14


F l em i ng ( 1 96 7 ) 13.4 17.7 -1 6 . 7 48 . 7 0.15
E
Mellor & Sep p ings ( 1 97 6 ) -4 . 0 1 4.4 - 36 . 4 30 . 7 0 . 28
Mod i f i ed Plank -1 9 .9 13.3 - 53 . 1 9.3 0 .22

Cowe l l ( 1 9 67 ) 34.5 19.9 3.4 85 . 5 0 .09


Lotz ( 1 97 4 ) 1 3.8 21 .3 -22 . 0 59 . 3 0 . 09
F
Mascheroni & Ca l v e lo ( 1 982 ) 15.5 1 6. 8 -1 6 . 6 54 . 1 0.1 6
de M i che l i s & Ca l ve l o ( 1 9 83 )
Pham ( 1 9 8 4 a ) 1 4 .7 7.2 0 .6 35 . 0 0 .54
Pham ( 1 98 4 c ) 18.3 48 . 8 -5 1 4
• 1 36 5 • 0 00

• • • continued
Tab l e 7 . 2 cont inued • • • 1 66

Group Method & Reference Mean Std Oe v 1 Min 1 Max 1 Corr 2


( %) (%) (%) (%) FDM

C l eland & Ear l e ( 1 982b ) 66 . 4 17.7 29 . 6 1 1 2.4 0.16


G Zar i tzky e t a l ( 1 9 8 2 ) 40 0 1 77 . 9 -64 . 2 208 . 0 -0 . 1 1
Hung & Thompson ( 1 983 ) 92 . 2 26 . 6 14.7 1 34 . 2 0 . 23
Pham ( 1 983 ) 53 . 6 18.4 21 .9 1 00 . 7 0.14

Calvelo ( 1 98 1 ) 0.0 6.0 -1 2 1


0 11. 1 0 . 66
H
Creed & James ( 1 98 1 ) 0.2 9.1 - 20 . 0 21 .9 0 . 52

Khatchaturov ( 1 95 8 ) 12.0 27 4
0 -45.7 86 . 6 -0 . 02
Gol ovk in et al ( 1 97 4 ) 7 .8 10 5
0 -1 3 .6 3 5 . 2 0 . 34
Schwartzberg ( 1 97 7 ) - 86 . 1 4.6 - 94 . 3 -71 . 8 -0 . 07
I
Church i l l & Gup ta ( 1 977 ) - 49 . 5 12.2 -73 . 5 -2 5 . 4 0 . 1 6
Levy ( 1 98 4 ) -35 . 8 9 .3 -6 4 . 8 - 1 7 . 7 0 . 30
Sastry ( 1 98 4 ) -57 . 4 10.7 -7 1 .8 -33 .0 0.31

1 Std Dev standard dev iat ion , Min


= m i n imum , Max
= max i mum.
=

2 Correlat ion coe ff ic ient ( r ) compared w i th the perc entage d i fferences


for the f in i t e d i fference method .
Pred ict ion of Thawi ng Times 1 67

Tab l e 7 . 3 Summary of Percentage D i f ferences Between Exper i mental


Thawi ng T i mes For Tylose Slabs , Inf i n i t e Cyl i nder s and
Spheres and Thaw i ng Times Calculated By the Best Present
Methods

Method 1 Data 2 Mean Std Dev 3 Max 3 Corr ..


( %) (%) (%) FDM

S , C , SP -1 . 3 4 .0 -9 . 5 9.1 1 . 00
FDM s -0 . 8 3.5 -6 . 3 6.8 1 . 00
c -2 . 2 2.4 -6. 8 3 .6 1 . 00
SP -0 . 9 5.5 -9 . 5 9.1 1 . 00

FEM f S , C , SP -0 . 9 4.4 -9 . 2 11 .4 0.94


s -2 . 0 3.6 -7 . 0 6.5 0 . 99
Eqs . ( 4 . 2 ) - ( 4 . 5 ) c -1 . 7 2.5 -6 . 0 4 .6 0 . 97
SP 1 .2 5. 8 -9 . 2 11 .4 0 . 99

FEM s S , C , SP -0 . 1 4.8 -7 . 6 13.1 0 . 88


s -2 . 2 3.6 -7 . 6 6.5 0 . 98
Eqs . ( 4 5 ) - ( 4 8 )
• • c -1 .0 2.6 -5 . 2 5.4 0 . 96
SP 2.9 5.9 -7 . 2 1 3. 1 0 . 98

S , C , SP 0. 1 5.7 -1 0 . 2 15.2 0 . 72
Eq . ( 1 . 2 ) s 0.4 6.0 -7 . 2 15.2 0. 71
c -1 . 6 3 .2 -1 . 1 5.2 0 . 66
SP 1 .5 6.9 -1 0 . 2 1 4 .6 0 . 76

S , C , SP -0 . 2 4.9 -9 . 2 1 2.0 0 . 80
Eq . ( 7 . 3 ) s -0 . 2 4 .5 -1 . 0 10.2 0. 72
c -1 .8 3 .3 -9 . 2 5.9 0 . 61
SP 1 .3 6.0 -9 . 0 12.0 0 . 89

S , C , SP 0.0 5.4 -1 2 . 6 14.3 0 . 80


Eq . ( 1 . 4 ) s 1 .7 6.0 -1 . 0 1 4 .3 0 . 17
c -1 . 1 3.1 -1 . 1 4.4 0 . 64
SP -0 . 5 6.2 -1 2 . 6 1 0.7 0 . 90

S , C , SP 0.3 5.6 -1 0 . 2 10.3 0 .76


Eq . ( 7 . 5 ) s 0. 1 5.2 -8 . 8 11 .1 0 . 72
c -1 . 3 3 .9 -8 . 8 5.9 0 . 69
SP 2.1 6.9 -1 0 . 2 1 4.8 0 . 79

1 FDM f i n i te d i f f erence method , FEM f


= ful l f i nite element method
s

formulat ion , FEM s s i mpl i f i ed f i n i te element method formul at ion .


=

2 S =slab ( 35 runs ) , C i n f i n i te cyl inder ( 3 4 runs ) , SP


= spher e ( 3 5
=

runs ) , data from Tab l es 6 . 2 , 6 . 3 , and 6 . 4 .


3 Std Dev standard dev i at ion , M i n minimum, Max max imum
= = s •

.. Correlat ion coeff i c ient ( r ) compared w i th the percentage d i fferenc es


for the f in i t e d i fferenc e method .
1 68

8 PR EDICTION OF FR EEZING AND THAW ING T IMES FOR


MU LT-D IMENSIONAL SHAPES BY NUMERICAL METHODS

The four th r esearch obj ec t i ve se t in Chap . 3 was to as sess the ac curacy


of Group I I ( numer ical ) methods for pred i c t i o n of freez ing and thawing
t i mes of irregular shap es . Chapter 8 dea ls w i th part o f the th i rd
obj ect ive ( assessment o f the accuracy o f pred i c t ion methods for regular
shapes ) and the fourth obj ec t ive . Resu l ts are r eport ed for nume r i ca l
methods app l i ed t o both regular and ir regular mult i-d imens ional shap es .

8. 1 I NTRODUCT ION

The deve lopment of numer ical methods that solve the govern ing par t ial
d i ffer ent ial equat io n for heat conduc t i on w i th temperatur e dependent
thermal properties for obj ects o f any two or three-d imens iona l geometry
was d i scussed in Chap . 4 . The fini te d i fferen ce me thod is l i mi t ed by
prac t ical cons iderat ions to regular shaped obj ects ( in f i n i t e rods ,
f in i t e cy l i nde rs and rectangular b r i ck s ) , wher eas the fini te e lement
method can tak e account of the more complex , irregular ,
mul t i - d i mens ional geometr ies as wel l as the regu l ar shapes. Thermal
property data used in all the numer i c a l calculations are l is ted in
Table 5 . 1 .

8.2 PR EDICTIONS FOR R EGU LAR SHA P ES

The only exper imental thawing data collected for mul t i -d i mens ional
regular shapes were those for Tylo se r ec tangular b r i ck s ( Tabl e 6 .5 ) .
The full f in ite element method formu lation ( Eqs . ( 4 . 2 ) to ( 4 . 5 ) ) was
not used as comp utat io n cos ts were too h ig h but the s impler f in i te
e lement method formulat ion ( Eqs . ( 4 . 5 ) t o ( 4 . 8 ) ) and the f in i te
d i fferenc e method were appl ied . Table 8 . 1 summar i ses the calcul a t ion
resu l t s .

Because of symmetry only an o ctant o f each rectangular brick was


cons i dered. Wi th in th i s octan t a 9 x9 x9 nod e gri d was u sed in f in i te
d ifference method calc ulat ions . The f in i te element method calculations
used a 7 x7 x7 node grid wi th 21 6 evenly s iz ed , 8 node , l in ear ,
Pred iction of Freez ing and Thaw ing T i mes 1 69

i soparame tr i c , b r i ck shaped e l ements . A f in er gr i d was not prac t i c ab l e


due t o computat ion c o s t and computer memory s i z e l imi ta t i ons .

Tab l e 8 . 1 shows that the s i mp l i f ie d f ini te e l ement method formu l a t ion


had s l i ghtly greater off-set of the mean pred iction error than the
pred ict ions for the f in i te d i fference method . As was the case for the
s imp le s lab , i n f in ite cyl i nder and sphere shap es th is d i fferenc e was
thought to be due to the re l a t i vely crude way in wh i ch var ia t i ons in
the thermal propert i es are incorporated into the s i mp l i f ied f i n i t e
element method . The three-d imens iona l geometry and the coa rse noda l
gr id use d , accentuated th is problem so the d i fferenc e between the
f in i t e d i f feren ce me thod and the s impl i f ie d f ini te e lemen t method is
grea ter than for slab s , i n f in i t e cylinders and spheres ( Tab le 7 . 3 ) .

Resu l t s for the ful l f in i t e element me thod formulat i on w ou l d be


expec ted to b e c l o s e t o the f i n i t e d i f ference method results ( Cl e land

et al 1 98 4 ) . For th irteen repres entat i ve r ectangular b r i c k fr eez i ng


runs the fu ll f i n i te element me thod results were calc ulated as we ll as
the results for the o th er two numer ical me thods . The mean pred i c t i on
errors for these runs were for the f i n i te d i f ference method ,
-5 . 7 %
- 8 . 2% for the s i mp l i f ie d f in i t e e lemen t method and - 6 . 3 % for the ful l
f in it e e lement method respec t i vely . The correlat io n coeffic ients for
the f in i t e e lemen t me thod perc entage d if f er enc es w i th the f ini te
d iffer ence perc entage d i fferenc es for these th ir teen runs were 0 . 9 6 and
0 . 99 for the s i mp l i f ie d and ful l formulat ions respec t i ve ly . Henc e , i f
the ful l f in i te element method had been used for the full se t o f data ,
the results wou l d have c lo sely matched the f in i t e d i f ference method
resul t s .

Pred icted temperature prof i les agreed c lo se ly w i th expe r i mentally


measured prof iles as i l lustrate d in F i g . 6 . 4 . A large par t o f the
d i fference between the pred icted and measured temperatures was probably
due to uncerta inty in thermocoupl e plac ement wh i ch was d i s cu ssed in
Sec . 5 . 7 . 4 . Both the f i n i t e d i fference method and the s impl i f ie d
f in i t e e lemen t method formu l a t ion pred icted surfac e temperatures that
osc i ll ated abou t the experimental values imme diately a fter the phase
change temperature r eg ion . Th i s was due to the d iscre te nature of the
numer ical methods and for the s imp l e f i n i te element method formulat ion
Pred i c t io n of Freez i ng and Thaw i ng T i mes

was magn i f ied by the c oarse element grid used for pred i c t i o ns .

Th e d if f er en ces between pred i c t ed and measure d thaw i ng t imes were not


sta t i s t i c ally correlated wi th any independent var iable in th e
exper i ments and d i splayed comparab l e accur acy wi th nume r i ca l
pred i c t i o ns for freez ing of Ty lose rectangu lar b r i ck s ( C leland & Earle
1 9 79 b , Cleland et al 1 984 ) . Fur the r the 9 5 % conf i dence bounds for the
nume r i cal me thods ar e s i mi lar in magni tude to the es t i mated
exper i men ta l unc erta inty ( Se c . 5 . 7 . 4 , Tab l e 1 2 . 2 ) . Th i s su gges t s that
for the f i n i te d i ff erence me thod and the full f i n i t e e lement me thod
formu l a t io n , mo s t of the d i f f er enc e b etween the pred i c t ions and
ex per imental values was due to exper imental uncerta inty and not errors
in the impleme nta t i on of the numer i cal methods themse l ves . That is ,
the t i me steps and th e space gr id used y i e lded resu l ts neg l i g ib ly
d i f feren t from those that would have been ob ta ined i f f i n er gr i d s had
been use d . The on ly s i g n i f icant inaccuracy a r i s ing from the numer ical
me thods themse l ves was tha t introduc ed by the crude th erma l property
est imat ion i n the simpler fi n i te element me thod formu la t io n ; th is
i nd icates that care shou l d be exe rc i sed i n us i ng it for rec tangu lar
br ick shapes .

8.3 PR ED ICTIONS FO R TWO- DIMEN SIONAL I RREGU LAR SHAPES

The ex per iment al fr eez ing and thawing t i me s for the two-d i mens ional
irr egu lar Tylo se shapes are g i ven in Tab l e 6 .6. Pred i c t ions were
carr ied ou t by both f in i te element me thod formulat ions us ing the gr ids
shown in F ig s . 5 . 1 5 to 5.21 . The gr i d s were pr epar ed from 8 node ,
quadrat i c , i soparame t r i c , rec tangular elements . These e lements allowed
curved b ound ar ies to b e appr ox imat ed by quad ratic func t io ns . The g r i ds
prepared ensured that at leas t f i ve elements ( 1 1 nodes ) e x i s ted betwe en
the thermo dynam i c c entr e and th e ou ts ide sur face of each ob j ec t , i n a l l
d ir ec t i o ns .

The results summar ised i n Table 8 . 1 show that both the finite element
formulat ions gave acc urate pred i c t ions of bo th fr eez i ng and thaw i ng
t imes ( w i th i n tolerances allowed for ex per i men tal error ) , for all the
shapes tested. The 95J conf i den ce b ounds for the p er centage
d i fferences of the full f in i te element formula t io n of -9 .7% to 1 0 .3%
Pr ed i c t i on of Fr eez ing and Thawing Times 171

are of s i mi l ar magni t ude to the es t i mated e xpe r i mental error


( Sec . 5 . 8 . 4 ) ind icat i ng that mo st of the d i f ferenc e was ex pla inab le by
the ex per imen ta l unc e rta inty . Aga i n , the ful l f i n i te e le me nt method
formu lat i o n used suff i c iently sma ll space and t i me i n tervals that
numer ical app rox imat ion error s wer e neg l i g i bl e , bu t th e s i mp l e f in i te
element me thod formu l a t i o n did i nt roduc e s i g n i f icant i nac curacy by
c rude therma l pr ope r t y es t imat ion .

Pr ed icted tempera ture profi les ( F i g s . 6 . 5 to 6 . 1 2 ) genera l ly compared


we l l wi th those measured exper i mental ly . D i fferen ces between pred i c ted
and measured tempera tures may have ar isen due to uncert ainty in
thermo coupl e placement for a l l the thermocouples . Also th e pos sib i l i t y
of air v o i d s , poor contac t and therefore var i at ions in the loc a l
surface heat tr ansfer coe ff i c ient mean t that measured t empe ratures a t
points on o r near the sur face had s i g n i f icant uncer ta int i es assoc iated
wi th them . Th e sur fac e temperatures pred i c t ed b y th e s i mpl if ie d f in i te
element me thod formulat ion osc i l lated s i gn i f icant ly compared w i th the
pred i c t ions for the ful l f i ni te e leme nt method formu l a t io n . Aga in th i s
d ifference was a t t r i b uted t o the crude thermal p roperty es t i ma t i o n in
th e s i mp l e formulat ion and the relat i ve l y sma l l numb er o f quadrat i c
e lements used i n the gr i d s for the pred ict ions .

8.4 PR ED ICT IONS FOR THREE-DIMEN SIONAL IRREGU LAR SHA PES

Tab l e 8 . 2 summar i ses the perc ent age d i f fer en ces between ex p er i menta l
freez i ng and thawing t i mes and th e pred i c t i o ns by the t wo f i n i te
e l emen t me thod formu l a t i ons for the three - d imens ional i r r egula r shapes .

The grid used for the f ir s t shap e ( th e fru stum of a sq uare pyram i d ) is
shown in Fig . 5 . 2 4 . Becau se of the s t ra ight s i des it is b as ica l l y the
same 7 x7 x7 node gr i d use d for the rec t angular bri ck pre d i c t ions , but
was d i s torted to the pyram i d d i mens ions . Be cau se of t he
three-d imens ional i r r eg ular i t y a ful l quadrant of the shape was
mode l l e d . Consequ en t l y th e grid was r a ther coarse wi th onl y three
l inear e lements between surface and centre a long the height of the
pyram i d . The exp er i menta l procedur e for the pyram i d was iden tical i n
a l l respec ts t o that u se d for the r e c t angular b ri ck shape so it was
cons i dered that the ex p er imenta l cond i t io ns wer e control l ed and
Pr ed i c t ion of Fr eez ing and Thawing T i me s 172

measur e d t o th e same leve l o f prec i s i on as in the rect angu l ar br i ck


exper i ments , and that the same sur face heat transfer coe f f i c ient
app l i ed .

The p re d i c t ions have s i mi lar spread t o those for the rec t angular b r i ck
freez i ng and tha w i ng but the mean was off se t by ab ou t 1 0 % . Th e l ow
standard dev iat ion of th e pred i c t i on er ror s suggests that th e prob lem
is sy stema t i c , and no t random. Two po s s ible r easo ns are : ( a ) th e
coa rse gr i d lead ing to s i gn i f i cant pred ict ion inaccuracy and ( b) th e
assump t i on tha t h was the same for th e pyramid and rect angu l ar b r i ck
shap es may not hold .

The amount of offse t was greater for the freez ing runs where B i was low
sugges t i ng that th e second r eason was an important con tr i bu tor
accord ing to the cr i te r i a of Cleland & Ear le ( 1 9 8 4 a ) .

In th e numer i ca l ca l cu l at i ons , freez i ng and thaw i ng were cons i dered


comp l e te when the slowe st coo l i ng or heat ing node reached the des ired
f ina l temperatur e . I f the true thermodynami c c entre of the ob j e c t does
not c o i nc ide w i th a nod e , at th is t i me , th en the obj e c t is not
comp le te ly fr ozen or thawed . Under-pred i c t io n of the t ime for t he
comp lete process w i ll r esult from th is effect . The effec t i s sma l lest
for f in e f ini te e l emen t gr ids , but for coa rse gr i d s t he ext r a t ime
cou ld be sub s t an t i al . Further , l in ear elements mag n i fy the problem
b ecau se th ey use l i ne ar appr ox i ma t io n funct ions betwe en nodes . The
true temp erature prof i le and any loc al max i mum or m i n i mum values of the
t emperatur e wi th in each e l emen t canno t be mode l l e d . Quad r at i c and
h igher order e lements a llow mo re acc urate ap prox ima t i ons to the
temperatur e prof i l e to be ma de and there for e give more ac curate
pred i c t io ns .

F i n i t e e lement method calc ulat ions for the pyramid shape were a ffec ted
in the manner j us t descr ibed, as a coar se gr i d o f l ine ar e l ements was
use d . Also the pos i t ion of the thermod ynam i c c en tre could not be
accurately determined p r ior to grid p reparat ion so the true
thermod ynamic centre was not necessar i ly c lose to a node . Therefore
the ob s erved under-pred i c t ion of freez i ng and thaw i ng t i mes was not
unex pec t e d . Pred i cted t emperatu r e p ro f i les were a dequate cons ider i ng
Pr ed i c t ion of Fr eez ing and Thawing T i me s 1 73

these prob l ems wi th the coa rse gr i d and the e xper imen ta l unc ert a i nty i n
thermo couple pos i tioni ng ( F i g . 6 . 1 3 ) .

Th e o th er three shapes u sed wer e a spher e , an egg and a f i sh shap e .


The sphere runs were use d t o prov ide informa t i o n ab ou t the surface heat
t r ansfer coe f f i c i en t for the o the r two three-d i mens iona l shap es mou l ded
i n p las t i c . The sphere is a regular shape and can be mod e l l e d by a
one- or two-d i mens iona l gr i d rather than three-d imens i onal ly if the
ax i symme t r i c al formulat ion of the finite eleme nt me thod is use d .
Howe v er mode l l ing th e sphere wi th a thr ee-d imens i ona l g r i d al low ed some
mea sure of the ac curacy of the mo dell ing of c urved surfac es by the
f i ni te e lemen t me thod to be made . The thr ee d i fferen t g r i d s shown i n
F i g . 5 . 25 we re used : (a) an 11 nod e , 10 l i near elemen t ,
one-d imens iona l grid ; ( b ) a two-d i mens iona l gr id mode l l i ng a qua drant
of the cross-se c t ion (a c irc l e ) w i th 21 , e ight node , qua dra t i c ,
i s oparame tr i c , e l ements and (c) a 208 node , 27 e lement ,
th ree-d imens io nal g r i d mod e l l ing an o ctant of the sphere . Due to the
l i mi t a t ions in comput a t i on power ava i l ab le and th e h igh data
prepara t ion t i me , the three-d imens ional g r i d was coarse w i th only thre e
2 0 node , quad rat i c , i s oparame tr i c el ements between t h e sur fac e and the
centre . These wer e arranged in three expand ing she l ls so that the
elemen ts were a l l appr ox i mate ly the same s iz e and none wer e t oo h i gh l y
d i storted from the b a s i c r eg ular rectangular b r i ck shap e .

The resu l ts generated for th e s ph er e shape us i ng b o th the f i ner one­


and two-d imens ional g r i d s w i th the full f in i te element formu l a t i o n were
very s i mi l ar . Th e coar ser thr ee-d i mens ional gr i d resu lts were
sign i f icantly less ac c urate . Th i s was a t t r i b uted to th e coarse grid
rathe r than incorrect i mp lementat ion of the three-d i mens iona l p ar t o f
the p rog ram ( Se c . 4 . 2 . 3 ) . The thre e quadra t i c elements betwe en the
c entre and sur fac e mea nt that the tempe rature p rof i l es c ou l d be
modelled only approx i mat ely . The pos i tion of the thermo dynamic c entre
was k nown exac t ly for the s phere , and quadrati c e lemen ts wer e u sed .
Therefore the problem with the coarse gr i d was less acute than for th e
pyrami d shape pred i c t ions and the amount o f under-pred i c t i on for t he
coarser thr ee-d i mens ional grid , compar ed w i th the f in er one- and
t wo-d i mens ional g r i d s , was less than the under-pred i c t io n for t he
pyram i d shape .
Pred i c t ion of Fr eez ing and Thawing Ti me s 174

For th e sphere shape th e s i mp l i f ied f in i te e lemen t method formu l a t i on


gave comparab l e re sults with the fu l l formu l a t i o n , but wh ere the
e l emen t s iz e was l arge r in the two- and thr ee-di men s ional gr i d s the
resu l t s had a l arger var i a t ion a t t r ib uted to th e less r ep resen t at i ve
incorporat ion of thermal p roper t i es i n to the f i n i te e lement scheme as
we l l as the coar se gr i d . For bo th f in i t e e l ement methods th e fr eez i ng
and thawing t i me pred i c t ion accuracy as well as the pred ict ion of .
temp era ture p rof i l es was s i m i l ar to that ach i eved for th e ful l set o f
sphere thaw i ng data analy se d in Sec . 1 . 2 . Th e experiment al techn iques
wer e s i m i l ar , exc ep t for u se of the p las t i c mould ing , so s i m i l ar
pred ict ion acc uracy was expected .

The egg shap e has one axis of rota t i onal symme try so it cou ld be
mode l l ed b y ei the r a two- o r thr ee - d i mens iona l gr i d . Bo th the f i n ite
e l emen t method gr ids used are shown in Fig . 5 . 26 . The two-d i mens ional
gr i d i s s i mi lar to that u se for th e sphe re shape exc ep t that it was
d i storted to f i t the elliptical shape and a ha lf , r ather than a
qua rte r , of the cr oss - sect ional prof i l e was mode l le d . The
three-d i mens iona l g r i d used was a lso a d istorted dou ble vers i o n of
th e sphere three-d imens iona l gr i d . Th e fu l l qu adran t need ed was
desc r i b ed by 5 4 twenty node , qua dra t i c , i soparame t r i c , r ec t ang ular
e lemen t s . The f i sh shape thoug h irr egu l ar in a l l three d imen s ions was
c lose ly approx imated by a shap e w i th two p lanes of symme try so on ly a
quadran t was mode l l e d . The g r i d i s shown in F i g . 5 . 27 . I t i s th e same
bas i c thr ee-d imens ional gr i d used for the eg g shape but the gr i d po i n t
coord inat e s , espec i a l ly f o r the sur fac e nodes were mod i f i ed t o f i t t he
ac tual shap e .

For the egg and f ish shapes it was assumed that the surface he at
tranfer coeff i c ient was the same as that me asur ed on the mou ld e d
p l as t i c s pheres ( Sec . 5 . 9 . 3 ) . Th e inaccuracy introduc e d by th i s
assump t io n was no t eas i ly asse s se d . Var i a t io ns i n p last i c th i ck ness
and the presenc e o f a ir vo i d s and poor er or be t t er surface con tac t at
the surfac e of these obj e c t s may have contr i b ut ed to the unc ert aint i e s .
Th i s extra e xp er i men ta l unc erta inty , p lus the unc ert a in t y d ue to
d if f i cu l t i es in measur i ng and desc r ib ing these shap es mathemat i ca l ly ,
a s wel l as appr ox i ma t i ng th e shape wi th a coar se f in i t e e lemen t method
g r i d , meant that poorer pred i c t i o ns were expec t ed for the egg and f ish
Pr ed i c t ion of Fr eez ing and Thaw i ng T i me s 1 75

shapes tha n for th e s phere and pyrami d shapes .

For the eg g shap e the me an pred ict ion error was c lose to zero so th e
assump t ion that the sur face he at transf er coeff ic ien t was th e same as
i t was for the sphere se emed reasonab l e . The h igher s t andard de v iat ion
of th e pred i c t ions for th e eg g shape compared w i th the py ram i d and
sphere shapes was ma inly th e result of the ex tra uncerta inty in the
exper i ment s rather than unc ert a inty due to th e implement at i on of the
f i n i t e e lement method . If a f i ner gr id ha d been use d , the f in it e
e l emen t method wou l d hav e introduced no s i gn i f ic an t pred i c t ion
inacc uracy .

Pr ed i c ted temp erature profi les for th e egg shape ( F i g . 6 . 1 6 ) we re a ls o


qu i te acc urate and reflec ted th e accuracy of the freez i ng an d thaw ing
t i me pred ic t i on s . Temperature pro f i les for some surface pos i t io ns may
have been affec t ed by any loca l i sed v ar i a t i on in h. Pr ed i c ted
temp erature prof i l e s were affected by the coarse three-d imens ional
g r i d s u s ed , and there for e d i sp l ayed some o sc i l l atory behav i our . As
exp ected , th is was part i c ularly obviou s for the s i mp l e f i n i te e lement
me thod formu l a t ion pred i c t io ns , for pos i t ions near the ob j ec t sur f ace
and at temp er atures close to the phase change temp erature range
( C le l and et a l 1 984 , Sec . 4 . 2 . 1 ) .

For the f i sh shap e the pred i c t i o ns were cons i s tently h igh by 20% to 4 0 %
b u t th e s t andard dev iat io n o f pred i c t ion errors was compa r ab l e w i th
that for th e egg shap e . The off - se t mean was probably caused by
sy stemat i c err or in the surfac e heat transfer coeff ic i en t . Th e surface
heat trans fer coeff i c ient may not have b een th e same as that for th e
sph er e du e to : ( a ) a th inner plas t i c th i ckness ar i s i ng dur i ng mou ld i ng
and/or ( b ) better therma l conta ct between the Ty los e and the plas t i c
moul d for the f i sh shape than the spher e . As d i scussed in Sec . 5 . 9 . 4
the se factors cou ld not no t be quantat i v e ly assessed as i t was no t
poss ible to measure h d ir ec t ly for the f i sh shape . Howe ver ,
measurement of th e plast i c th ickness showed variat ions of up to 1 5 % of
the average to ta l th i cknes s ove r the surface of th e f i sh shape and
d if ferenc es of up to 1 0% i n the th ick ness b e t ween the f i sh and sphere
shapes . It was unknown whe ther these ob served va r iat ions were
suff i c i en t ly large to cause the d i fferences betwe en the pred icted and
Pred i c t ion of Fr eez i ng and Tha w i ng Ti mes 1 76

me asured fr eez ing and thaw i ng t i me s . Because the f in i te e lement me thod


d i splayed ac curat e pred i c t ions for o the r r egu l ar and irr egul ar shap es
it was cons i dered unl i k ely that these d i f ferenc es were the result of
de f i c i enc i es in the f in i t e e lemen t me thod or p rogram .

Pred ic ted t emp erature pro f i l es for the fi sh shap es ( F ig . 6 . 1 5 )


refl ect ed the acc uracy o f the pred i c te d fr eez i ng and thaw i ng t i mes and
p roblems w i th the coarse element g r i d used . The f in i te e lement me thod
cons i s ten t l y under-pred i c ted th e changes in th e temperature at a l l
pos i t i ons i n the f i sh sha p e . Temperature prof i les for surface
pos i t io ns would have been affected by any local i sed var i a t ion in the
sur fac e hea t transfer coe ffic ient as well as th e pos s ib le systemat i c
error in the h valu e , b u t th e pred i c ted values showed tr ends co n s i stent
w i th a sy stema t i c underpre d i c t ion o f h.

8 .5 F IN ITE ELEM ENT METHO D USER GU ID ELI NES

Some gu i de l ines for appl i ca t ion of th e f inite e l eme nt method were


establ i shed in Sec . 4 . 2 . In the l ight of the tes t i ng ag a inst
ex peri men tal da ta i t i s po s s i b l e to ref ine these. Adher ence to the
guidel ines w i l l lead to p re d i c t i o ns in wh ich the er ror a r i s ing from the
numer ical appr ox i mat ions is negl ig i b l e comp ar ed wi th e xper imenta l and
data unc er t a i n ty . The gu ide l i nes are :
( a ) to use at least 8 to 1 0 nodes and a t l e ast 5 e lements b et we en the
ther modynam i c c entre and the sur fac e ,
( b) to use the full f i n it e e lement formulat io n if compu t at io nal
l i mi tat ions a l low ,
( c ) to ac c urat e ly pred i c t th e pos i tion of the thermodynami c centre
p r ior to gr i d prepa r a t io n so that a nod e can be p os i t i oned
accura tely near i t , and
( d ) to use suff i c iently sma ll t i me steps that the heat balance a grees
to wi th i n 2% ( Sec . 4 . 2 . 2 ) .
In a lmo s t all s i tuat i o ns , u s ing more r e f i n ed gr i d s and smaller t ime
s t ep s than tho se sugges ted h er e , wi l l on l y g i v e smal l i nc r eases i n
accuracy and re l i ab i l i ty o f the p red i c t ions and will probably not
j u s t i fy the i n crease in computat ional cos ts .
Pr ed i c t ion of Fr eez ing and Thawi ng T i me s 1 77

8 .6 SUMMARY

The f i n i te d i f ference me thod use d for reg ular two- and


three-d i mens io na l shapes pred i cted thaw i ng ti mes as ac curate l y as i t
pred icted freez ing t i mes . Th e me thod cou ld be imp lemented w i th
n egl ig ible inacc uracy ar is i ng fr om the numer ical app rox i mat ions u sed .

The sp reads of the nume r i c al me thod pred ict ions were low for a l l da t a ,
b u t mean pred i c t ion error s for some thr ee-d imensiona l irr egula r shapes
were of f-se t from z ero. Th is ind icates tha t systema t i c data er ror s ,
probab ly i n es t i ma t i on of surface heat t r ansfer coe ff ic i ent s , were
impor t ant compared wi th random exp er imental unce rtainty . Pr ed icted
tempera tur e prof i l es conf i rme d these t r end s .

The i mp lemen tations o f the f i n i t e element me thod used for both regular
and irregular mu l t i -d i mens iona l sh ap es do not lead to sign i f i cant
inacc uracy in p red i c t ions p rov ided the gu i d e l i nes in Se c . 8 . 6 ar e
fo llowed . U se of coarse gr i ds and /or the s i mpl i f i e d f in i t e e l ement
method formulat ion d id lead to s i g n i f i cant inac curracy in the
pred ic t ions and there fore wor sened overa ll agreemen t w i th ex pe r i mental
freez ing and thaw ing dat a .

Pr ac t i cal contra int s o n comp uta t i o n power me ans that use r s of the
f in i t e e leme nt me thod w i l l have to wor k c l ose to th e l i mi ts of g r i d and
method re f inement for wh ich p red i c t ion me thod uncertainty is
s ign i f ican t , so ca re mus t b e ex erc i sed in u s i ng these methods .
Pr ed ict ion of Fr e ez ing and Thawi ng T i mes 178

Tab l e 8 . 1 Summary o f Perc entage D i ffer enc es Bet ween E xperi menta l
Freez i ng and Thaw i ng Ti mes For Ty l ose Mul t i - D i mens i onal
Shapes and Freez i ng and Thaw i ng T i mes Ca lcula ted By Numer i ca l
Methods

Shape Type 1 Re ference Method 2 Mean St d Dev 3 Max 3


( %) (% ) (%)

C l e land & FDM -4 . 4 6.6 -20 . 7 9.8


F Ear l e ( 1 9 79 b)
Rec tangular ( 7 2 runs ) -5 . 9 7.1 -2 4 . 1 9.8
Br i ck
Table 6 . 5 FDM 0.6 3.1 -6.0 7.9
T
( 68 runs ) FEM s -6 . 3 3.4 -1 2 . 9 1 .2

Tab l e 6 . 6 0.3 5. 1 -1 1 . 4 1 1 .3
F ,T
( 83 runs ) 4 .2 6. 1 -9 . 1 21 .9

Two­ Tab le 6 . 6 -1 . o 4.5 - 1 1 4


• 7.6
D i mensional F
Irregular ( 4 2 runs ) 2.6 4.8 -9 . 1 1 2 .0

Tab le 6 . 6 1 .5 5.5 -7 . 1 1 1 .3
T
( 4 1 run s ) 5.8 7 .0 -6 . 7 21 . 9

1 F = freez ing , T = thaw i ng .


2 FDM = f i n i t e d i f ference me thod , FEM = full f i n i te element me thod
f
formulat ion ( Eqs . ( 4 . 2 ) to ( 4 . 5 ) ) , FEM s = s i mp l i f ied f i n i t e e l ement
me thod formulat ion ( Eqs . ( 4 .5 ) to ( 4 . 8 ) ) .
3 Std D e v = s t andard dev iat ion, M i n = m i n i mum, Max max i mum.
Pr ed i c t ion of Fr eez ing and Thawing T i mes 179

Tab l e 8 . 2 Summary of Perc ent age D i f fer enc es Bet we en Exper i menta l
Freez i ng and Thaw i ng T i mes For Ty l ose Three-D i mens i ona l
I r r egular Shapes and Freez i ng and Thaw i ng T i me s Ca l cul ated By
the F i n i te El ement Method

Shape Gr i d 1 Method 2 Mea n Std De v 1 Min 1


( %) ( %) ( %)

Pyramid -1 0.3 6 .7 -18.2 -2.0


3D
( 6 runs ) -1 4 . 5 4 .8 -20 . 2 -8 . 9

-5 .5 6 .8 -1 2 . 1 5.3
3D
-1 2• 12.7 -1 6 . 1 13.9

Sphere 0.6 7 .0 -8.7 13.9


20
( 6 runs ) 7.7 8.3 -1 . 6 18.2

0.4 6.7 -7 . 7 9 .6
10
2 .6 6.5 -4 .9 1 1 .6

5.2 1 3.0 -1 2 . 9 25 . 1
3D
Egg 9.8 14.9 -9 . 4 28 . 8
( 6 runs )
3 .7 1 3 .8 -1 6.1 24 . 7
2D
16.4 1 8 .7 -4 . 1 48 . 4

F i sh 35 . 3 8.9 20 . 8 44.2
3D
( 6 runs ) 33 . 5 1 6.3 8.6 58 . 4

1 1D = o ne-d i mens i o nal , 20 two-d imens ional , 3D three-d imens ional .


2 FEM
f full f i n i t e elemen t method formulat io n ( Eqs . ( 4 . 2 ) to ( 4 . 5 ) ) ,
EM
F s s i mp l i f ied f in i te elemen t me thod formulat io n ( Eqs . ( 4 .5 ) to
( 4 .8 ) ) .
3 Std De v standard de v iat ion, Mi n min imum , Max max i mum.
1 80

9 PREDICTION OF FREEZING AND THAWING T IMES FOR MULTI-DIM ENSIONAL


R EGU LAR SHAPES BY SIMPLE M ETHODS

I n Chap ter 9 Group I ( s impl e ) me thods for regu l ar mul t i - d i mens i onal
shapes are cons i dered . Th i s comp l etes the ana lysis requ ired for the
th ird research obj ec t i ve set in Chap . 3 .

9.1 ANALYT ICAL TR EATMENT OF TH E EFFECT OF GEOM ETRY

For s l ab s , i n f i n i te cy l i nders and spheres the 6 : 3 : 2 rat io of freez i ng


and thawi ng t i mes , wh i ch ar i s es from the value of 2V/ AD , was
success f u l l y used to descr ibe the effec t of the geometry so that s i mp l e
pred i c t i on me thods could be made un i versal for a l l three b as i c shapes .
Rutov ( 1 936 ) and Mott ( 1 9 64 ) proposed the use of the same 2 V / AD ratio
for a l l shap es . The pred i c t ion equa t ion becomes :
2V
t t slab ( 9. 1 )
AD
where t free z i ng or thaw i ng t i me for the mul t i-d imens ional
obj ect of charac t er i s t i c d imens ion D (s)
freez i ng or thawi ng t i me for a s l ab of
th i ck ness D under equ i val ent cond i t ions
A = surface area of the obj ec t
V volume o f the obj ect
D = charac ter i s t i c d imens ion of the obj e c t (m)
l ength o f shortest d imens i on f o r r egul ar
shaped obj ec ts .
For two- and three-d imen s ional regular shapes ( infinite rods , finite
cyl i nders and rectangular b r i ck s ) an al t erna t i ve approach was used by
P lank ( 1 9 4 1 ) and Shamsundar ( 1 982 ) . They made the quasi -st eady s tate
assump t i on and assumed both 4 par t i cular shape for the phase change
front and that the surface was isothermal w i th respect to pos i t i on .
Th i s l ed to geome tr i c factors to mod ify the bas i c s l ab formulae that
t end to 2 V / AD as B i +O but i nc luded a smal l correct i on to the i n t ernal
conduc t ion t e rm , in the r esu l t i ng analy t i ca l formu l a , for o ther B i
val ue s .
Pred i c t i on of Freez i ng and Thaw i ng T imes 181

For a l l two- and three-d imens ional sha p es ( apart from inf i n i te
cy l i nders and spheres ) the sur face i s not i sothermal w i th respect to
pos i t ion unl ess B i +O or B i � � . I n add it ion the phase change front shape
var i es wi th t i me . C onsequent ly the true geome t r i c factors for
mu l t i - d imens iona l shap es ar e in rea l i ty h i ghly dependent on the B lot
number . As B i +O the ratio of 2V/ AD ho lds for a l l shapes . For h igher
Bi values , some parts of the surface ar e more effec t i ve in transf err i ng
heat to or from the centre than others . There fore none of the
analyt i ca l developments based onl y on the area to volume r at io are
successful for freez i ng ( Cl eland & Ear le 1 98 2 b ) . The comp l ex i t y of the
heat transfer in mul ti - d i mens ional obj ects is such that an accurate
geometr i c fac tor is un l i k ely to b e der i ved analy t i ca lly .

9.2 FEAS I BL E GEOMETRIC FACTORS

Al terna t i ve geomet r i c fac tors have been developed , par t ly by f i tt ing to


exper imental data and par tly by cons i der i ng some known l im i t i ng cases .

Cl eland & Earle ( 1 982 b ) de f i ned the equ ivalent heat transfer
d i mens iona l i ty , EHTD , as :
t slab
EHTD = (9.2)
t
The formul a for rec tangu l ar b r i ck shapes :

Bi [ 0 . 625 + 0 . 6 251
(9.3)
+
Bi + 2 � a� J
where sl : rat io of second longest to shor test s i de l eng th
.. D y iD x
61 a rat i o of l ongest to shor test s i de l eng th
- D z /D x
• shor t est s i de l eng th (m)
• second longest s i d e l eng th ( m)
• longest s i de l eng th (m)
was deve loped by f i t t i ng exper imen tal data w i th B i in the range 0 . 5 to
22 . 0 . I t does not conform to the l im i t i ng case that EHTD•AD/2V a s B i +O
but does fit pub l i shed exper i mental data for rectangular b r i ck s
6 62
1
undergo i ng fr�e z i ng . I t covers ranges of Bi , and t yp ically
encount er ed in prac t i ce . Used i n conj unc t ion wi th the freez i ng t i me
pred i c t ion formula for slabs due to C l e land & Earle ( 1 982 b ) it
Pred i c t ion of Fr eez i ng and Tha w i ng T imes 1 82

pred i ct ed the ex per iment al Tylose rec tangu lar br i ck free z i ng data of
C l e l and & Ear l e ( 1 97 9 b ) wi th percentage d i fferenc es wi th a mean of
- 1 . 2% ±1 1 . 4% a t the 95% level of con f i dence .

Pham ( 1 98 4 b ) used the 2V/AD geometr i c factor but mod i f i ed the i n t ernal
conduc t i o n term in the s i mpl e pred i c t i on formula based on Plank ' s s l ab
equat i on by a factor ca lled the " mean conduc t i ng path leng th" , MC P .
MC P is based on the same p r i nc iple as the corr ect ions der i ved
ana ly t i ca l l y by P l ank ( 1 9 4 1 ) and Shamsundar ( 1 982 ) but b ecause it was
developed based on fit to ex per imenta l data it i nc ludes compensat i on
for ef fects such as the non- i so thermal sur fac e , that P l ank and
Shamsundar i gnored . For rec tangu lar b r i ck s the formula deve loped to

[(-1
ca lculate MC P was :
MC P
-- - 1
D
+
[( 1 .5/B, 1 )
-4 +

e1
+
1
-H1 + -)
s2
4
Bi
I] l
-4 0 . 25
(9.4)

For f i n i te cyl inders , an equ ivalent rec tangular b r i c k shape was de f i ned
so that Eq . ( 9 . 4 ) cou ld be used ( Pham 1 984 c ) . Equa t i on ( 9 . 4 ) , comb ined
w i th the slab pred i c t ion me thod of Pham ( 1 9 84 a ) , pred ic ted the Tylose
rec tangular b r i ck freez ing data ( Cl eland & Earle 1 97 9 b ) wi th 95%
con f i dence bounds of - 1 3 . 0 % to 1 3 . 0% and a mean p�ed i c t ion error c l ose
to zero , wh ich is s i m i l ar to the results for EHTD ( Pham 1 98 4 b ) .

MCP and EHTD are essen t i a l l y equ i val ent geome t r i c correc t ions to the
bas i c slab pred i c t ion methods but are i ncorporated into the s i mple
formulae in d i fferent manner s . EHTD i s a d i v i sor and can b e app l ied to
any slab pred i c t ion method by rearrangement of Eq. ( 9 . 2 ) to g i ve :
t slab
t = (9.5 )
EHTD
MC P can be d irectly incor porated into s i mple pred i c t ion formulae where
contr ibut i ons due to ex t ernal and internal r es istance to heat transfer

t
V
[
can b e separated ( such as Pl ank ' s equat ion ) , by the re lat ionsh i p :

External R e s i stance + I nt ernal Res i s tanc e MC P


] (9.6)
« A Term Term -o-
A mor e general form wh i ch allows it to be app l i e d to any slab

[ ] [ -�
pred i c t ion me thod i s :
2V B i MC P Bi
t •
t s l ab - 1 + - -- I 1 + (9.7)
AD 4 D 4
Pred i c t ion of Freez ing and Thaw ing T i mes 1 83

[1 + ] [1 + - ]
The values for the two var i ab l es can be related by the equat ions :
AD Bi B i MC P
EHTD = - I -- (9.8)
2V 4 4 D

[
and

+1 1
J
MC P 4 AD 4
(9.9)
D
=

Bi 2V EHTD Bi
EHTD and MC P de f i ned by Eqns . ( 9 . 3 ) and (9.4) are both accurate
geome tr i c factors for the rec tangular brick shape wi th the intermed i ate
8 and B i va lues that occur in the rectangular b r i ck freez i ng da t a from
wh ich they were developed . For other cond i t i ons they may not be
accurate .

One pos s i b l e app roach for the current work was to seek to ex tend the
MC P and EHTD de f i n i t ions to a w i der range of geome t r i es and
env ironmental cond it ions to enhance the ir versat i l i ty . If th is was
succ essful then there wou ld be l i ttle bene f i t in seek i ng to develop
comp l etely d i f ferent geometr i c fac tors . The re fore the ap p roach taken
was to i nves t i gate the MC P and EHTD conc ep ts in deta i l , and then to
assess whether the ir l ack o f fit to data was suf f i c iently great to
j us t i fy cons i der i ng other a l t ernat i ve geome tr i c fac tor s .

I f both MC P and EHTD can be made accurate for a w i de range of


cond i t i ons and geometr i es then the cho ice be tween the two conc ep ts must
b e made by cons i der i ng other cr i t er i a such as :
( a ) concep tual understand i ng and s i mp l ic i t y
( b ) ease and accuracy of calculat ion for a wide range of obj e c t
geome t r i es .
Because the equa t ions to calculate MC P and EHTD , such as Eqs . ( 9 . 3 ) and
( 9 . 4 ) are par t l y emp i r i cally deve l oped i t i s poss i b l e that the s econd
cr i ter ion a l l ows a d i s t inction to be made .

9.3 VER I FICAT I ON OF THE EFFEC T OF ENVI ROMENTAL CONDIT I ONS


ON GEOMETRIC FAC TORS

As de f i ned by Eqs . ( 9 . 5 ) and ( 9 . 7 ) EHTD and MC P are both re lated to the


ratio of t slab / t . For i n f i n i t e cyl inders and spher es it was found
( Sec . 7 . 1 ) that the dependence of this rat io on B i , Ste and Pk was
suf f i c i en t l y weak for both freez i ng and thawing that the var i a t i on
Pred i c t ion of Freez ing and Thaw i ng Times 1 84

could b e i gnored . For other mul t i -d i mens i onal shapes the B i effect
wou ld b e expec ted to b e larger because the surface is not i sothermal
wi th r espect to pos i t i on and the degree of approach t o a constant
sur fac e temperature is Biot number dependen t . Where mos t of the
resi stance is int ernal ( B i � m ) , sur face area not d irectly adj ac ent to
the thermodynam i c centre w i l l have l i ttle effect on the change in
temperature at the centre . Conv ersely i f mos t of the res i stance is
ex ternal ( B i � ) . then a l l the sur face ar ea of the obj ect is equa lly
e f fec t i ve and no fur ther correc t i on to 2V/ AD i s needed .

Free z i ng and thaw i ng t i mes were calculated for d i fferent B i , Ste and Pk
values u s i ng the f i n i te d i ff erence and f i ni te el ement me thod s .
Calculat ions were made for a range of di fferent i n f i n i t e rod , f i n i t e
cy l i nder and rectangular b r i c k shap es as we l l as for the equ i va l en t
slab shape so tha t the dependence of t slab /t on Bi , Ste and Pk cou ld be
exam i ned .

The f i n i t e d i f f erence method resul ts for free z i ng and thaw ing of two
i nf i n i te rod shap es are shown in Tab le 9 . 1 . These are typ i cal for al l
the shapes , for both free z i ng and thaw ing , and for both the numer i ca l
methods used . For each of the shap es t ested the effect of changes i n
Ste and P k , due t o changes in T a and T in ' wer e far less than the e ff ec t
of changes i n B i for the ranges of cond i t i ons typ i ca l l y encountered i n
prac t i ce . The Biot number effec t must be tak en i n t o account but the
Ste and Pk effects are smal l . The Ste and Pk effects were most
impor tant for thaw i ng , that process changed t slab / t by less
and for
than ±5 % over the full range of Ste and Pk va lues typ ically encountered
in prac t i c e . The i n i t i al approach was to ignore the Pk and Ste
effec ts .

9.4 D EV ELO PMENT O F IM PROVED GEOMETR I C FACTORS FOR MU LT I- DIMEN SIONAL


R EGU LAR SHAPES

The analysis of the p rev ious sect i on conf irmed that for regular
mul t i - d i mens i onal shapes , accurate geomet r i c factors can p robab l y b e
defined solely a s func t i ons o f B i and parameters that descr i b e the
shape , wi thout i ntroduc i ng s i gn i f i cant error into the pred i c t i on
method :
Pred i c t i on of Freez ing and Thawi ng T i me s 1 85

Geometr i c Fac tor = f ( Bi , �1' �2 ) (9 . 1 0 )


The ex i s t i ng formu l ae for EHTD and MC P ( Eqs . ( 9 . 3 ) and (9.4)) w ere
curve- f i t ted to exper imental data for freez ing of rectangular b r i ck
shapes wh i ch were both l i mi ted in range of �1' �2 and B i , and a ffec ted
by exper imental error . To deve lop an improved emp i r i ca l geome t r i c
fac tor bet ter data were consi dered necessary . A n alternat ive data se t
that shows the geome try effect wi thout mask i ng by exper i me ntal
uncerta i nty was deve loped by using numer i ca l pred i c t ion methods for
regul ar mu l ti - d imens i onal shapes .

The numer i c a l ca lculat ions were carr i ed ou t over a wi der range of


cond i t ions than the pub l i shed exper imental dat a . Freez i ng t i mes were
calculated for fre e z i ng of the inf i n i te rod , fin ite cy l i nder and
rectangu l ar b r i ck shap es us i ng the f i n i te d i fferenc e me thod . B i ot
numb er was var i ed from 0 . 0 1 to 1 00 . 0 , and �1 and �2 from 1 . 0 to 10.0.
For the rect angu lar br i ck shape on ly the ten poss ible comb inat ions
der i ved from �1 and �2 values of 1 .0 , 1 . 582 , 2 . 50 and 4.0 were
i nv es t i gated due to l im i t a t ions in compu t a t i on r esourc es . The
equ ivalent slab shap e free z i ng t i mes were a lso ca lculated usi ng the
f i n i te d i ff erence me thod . Equat ions (9.2) and (9.9) were used to
calculate EHTD and MC P . Using the same pred ict ion method for both the
mul ti -d i mens ional shapes and the equ ivalent s l ab g i ves a sys tema t i c
error cance l l i ng e ffect s o that the results reflec t a lmos t ent i r e ly the
var i a t i ons in geome tr i c factor s . Cal cul at ions for some runs us i ng the
full f i n i te element method formulat ion gave i dent ical results to the
f i n i te d i f f erence me thod , so the ful l f i n i te e l ement me thod formul a t ion
was not used for the full set of data . As sl i gh t changes in the
geome t r i c fac tors can occur wi th env ironmental cond i t ions , all
calculat ions were made a t i ntermed iate Ste and Pk values to average
these e ffects . A ful l set of data was not calcul ated for tha w i ng as a
tr i a l of some rep resentat i ve calculat ions y i e lded very simi l a r MCP and
EHTD values to those calcul ated for freez i ng .

The numer i ca l l y calculated results are are tabulated i n App . C ( Tabl e


C . 1 ) . Apart from some random error introduced b y s l i ght i mprec i s i on i n
the numer i cal calcu l a t i ons , espec i a l ly a t extremes o f the range o f Bi ,
the r esul ts showed a number of cons is tent patterns. Each shape was
cons i dered to have thre e d imens ions , taken in order of s i z e . The f i r s t
Pred i c t i on o f Freez i ng and Tha w i ng T i mes 1 86

d imens ion was the character i s t i c d ime ns ion and had an e ffect equal to
that of a slab for i nf in i te rods , rect angul ar b r i cks and squat
cy l i nders ( f i n i t e cyl i nders w i th D r � D y ) , and an effect equal to that
of an i n f in i te cy l i nder for short cy l i nders ( f i n i t e cy l i nders w i th D r �
D y ) . The e ffec t of each of the other d imen s i ons on the rat io o f
t s lab / t f o r each shape cou l d b e separated . A number of trends were
ev i dent
( a ) the e ff ec t of the second d imens ion for infinite rods and
rectangular b r i cks was equal
( b ) the e ffect of the second and th ird d i mensi ons for the squat
cy l i nders ( cons i der i ng the th ickness as the f irst d i mens ion and the
d i ame t er as two equ ivalent d i mens ions ) were b oth the same , and were
equal to the e f fect of the second d i mens ion i n inf i n i te rods and
rec tangu lar bri ck s
( c ) the e ffec t of the th ird d imens ion i n short cy l i nders ( consi der i ng
the d i ame ter as the f i rst two d i mens ions and the hei ght or l ength
as the th ird ) was the same as the effec t of the th ird d i mens ion in
rec tangular b r i ck s , but was l ess than the effect of the second
d imens i on i n i n f i n i t e rods and rectangu lar b r i ck s .
The va lue of t slab / t changed wi th B i ot numb er . The va l ue o f B i at
wh ich the change with respect to B i was greatest depended on the rat io
o f d i mens ions :
( d ) as e 1 or e 2 i ncreased the change occurred at l ower Bi values
( e ) for h i gh e 1 and e 2 v a lues as B i ��, the values o f MC P and EHTD both
t ended to s l ab l i mi ts ( for inf ini te rod s , rec t angu lar b r i ck s and
squat c y l i nders ) or i n f in i t e cy l i nder l im i t s ( for sho r t cy l i nde r s )
( f ) for l ow e 1 and e 2 va lues as B i � � , the val ues of MC P and EHTD b oth
tend to some o ther l i mi ts wh i ch are frac t ions of the v a l u es as
B i +O .
( g ) as B i +O the values o f MC P and EHTD both depend only on 2 V / AD .

U s i ng non-l i near r egress ion of the numer i ca l data set , a numb er of


func t i ons were dev e l oped to calculate EHTD and MC P for all t h e regular
shap es . The s i mp l es t forms that gave r easonab l e pred i c t ion accuracy
wer e :
P r e d i c t ion o f Fr eez i ng and Thawi ng T imes 1 87

EHTD G1 + G2 [ r.,<e,l + (1 - f 2 < 81) )


]
0.73
?
]
81 1

+ G3 [ r.,<e,l 82
+ (1 - f 1 2( 82) )
0 . 50
� (9.1 1 )

and
MC P
( 2 - G1 ) f ( f 1 3 ( G2 , ) 0 . 63 8iO .
39
34 8� 43 - 1 )
+
1 .. 81) (1 .
D

0 39 22
+ G3 f 1 .. ( f 1 3 ( G3 , 8 2 ) ) 0 . 63 8i · (1 . 51 8;· - 1 )

+ [ 1 - f 1 .. ( (f 1 3 ( G2 , 8 1 ) + f 1 3 ( G3 , 8 2 ) ) I ( G2 G3 + 1) ) ] .

[
( G1 + G2/ 8 1 + G 3 / 8 2 )
(9.12)
(G1 + G2 0 . 7 3 / 8�·50 + G3 0 . 5 0 / 8�· 69 )

1 . 77 1 1 . 34 . 77
where f 1 ( S) 2 . 32/ 8 (8i + 2 . 32/ 81 )
2
92
f 1 3( G, 8) G 2 . 00/ 81 ·
3
f I .
. ( X) x/ ( 8i 1 . 4 + x )

and the v a l u es o f G 1 , G2 and G3 a r e g i ven i n Tab l e 9 . 2 .

The p r i nc ip l es beh ind the methods to calcu l ate EHTD and MC P wer e

deve l oped from the ab ove observed trends i n the t s l ab / t data . For the

MC P conc ep t the A D / 2 V r a t i o was i ncorporated i n d ependent l y ( Eq . ( 9 .7 ) )

whereas for the EHTD c o nc ep t i t was inc luded i n the geome t r i c fac tor

i t se l f .

The geome t r i c fac tor c a n b e i n fl ue nced by u p to thr e e d imens ions for

any shap e . F i r st l y , the e f f e c t of the f i r s t ( charac ter i s t i c ) d i me ns i on

i s added ( G1 for EHTD and 1 .0 for MC P ) . Wh ere the char ac t er i st i c

d i mens ion can b e measured i n mor e than one d i me ns i on , the e f f ec t of

each of the add i t i o n a l d imens i ons i s to i nc rease the v a l u e o f G 1 by

one . The data in Tab l e 9 . 2 i l l us t r ate th i s i dea . The e f fe c t s of each

o f the other d imen s i on s ar e then ca lcu lated and summed as needed .

As 8 i +O the e f fec t depends on the AD/ 2 V rat i o o n l y . For reg u l ar sha p e s

AD/ 2V 2 G 1 + G2 / S 1 +G 3 / 8 2 s o for EHTD the add i t ions ar e G2/ 8 1 and/or

G3/ S 2 , wh i le for MCP there i s z ero add i t i on a s the e ff ec t of AD / 2V has


a l r eady b e en i nc l uded .

As 8i +oo the e ffect of each d i mens i on i s l ess than that sugge s t ed by the

AD/ 2V r a t i o so the e f f e c t o f ea c h d imens i on is a frac t ion of G2/ S 1


P r ed i c t ion of Freez i ng and Tha w i ng T i mes 1 88

and /or G3/ e 2 for EHTD and is greater than zero for MC P . The frac t ional
values , wh i ch are d i fferent for each d i mens ion because the ef fec t of
the th ird d i mens i on is l ess than the effect of the second d imen s ion ,
are g i ven b y 0 . 7 3/ e � · 50 and 0 . 5 2 / e � · 6 9 respec t i v e ly for EHTD .

For MC P the f i nal t e rm ca lculates the mod i f i ca t ion t o MC P requ i r ed to


al ter the effec t
o f the AD/2V rat i o . I t i s based on the f ac t that as
B i � , MC P � EHTD ( B i � ) / EHTD ( B i � m ) .

For the EHTD concep t the exp ress ion is comp let ed by a B i ot nu mber
we igh t i ng func t ion that calculates the relat i ve con t r i b u t ion of the
Bi � and the B i � m t erms in the ef fect of each d imens ion . The equa t ion
can b e wr i t ten in general terms as :
EHTD = XI ( EHTD Bi � ) + ( 1 - x
l ) ( EHTD Bi � m ) (9.13)
where x
l Biot numb er we ight ing func t i on ( 0 � x
l � 1). The Bi va lue
at wh ich the max imum rate of change , from one t erm to the o ther , occurs
is dependent on the re lat i ve s i z e of tha t d imens ion compared w i th the
charac t er i s t i c d i me ns ion .

For the MC P concep t the transi t ion i s more d i f f i cu l t to calculate as


MCP goes through a max i mum va lue b e tween the two B io t number ex tremes .
Ther e fore the express i o n inc ludes an extra Blot number correct ion as
well as a B i we igh t i ng func t ion s im i lar to that used for EHTD . Wr i t t en
i n general terms :
MC P = X 2 MC P B i � m + ( 1 X 2 ) ( EHTD Bi � I EHTD Bi � m )
-
(9.1 4)
where X 2 = B l o t numb e r we i gh t i ng funct ion ( 0 � X 2 � 1 ) and MC P B i � m is a
func t ion i nc lud i ng a t e rm that increases as B i i ncreases .

Tab l e 9 . 3 shows the adequacy of the f i t to the numer i ca l t s l ab / t dat a


of ( a) the old EHTD formula ( Eq . ( 9 . 3 ) ) , ( b ) the new EHT D formu l a
( Eq . ( 9 . 1 1 ) ) , ( c ) the old MC P formula ( Eq . ( 9 . 4 ) ) and ( d ) the new MC P
formula ( Eq . ( 9 . 1 2 ) ) . The new vers ions of EHTD and MC P are s i g n i f i cant
imp rov ements over the older ver s i ons . Th is was expec ted b ecause
Eqs . ( 9 . 3 ) and ( 9 . 4 ) were b o th developed from data for the r ectangular
on ly and for a l imi t ed range of Bi , e 1 and e 2 v a lues .
b r i ck shap e The
inaccuracy in the improved MC P and EHTD formulae ( ab ou t ±3% at the 95%
l evel of con f i denc e ) is l ikely to be insign i f i can t compared wi th
exper imental uncerta int i es . There fore the introduc t ion o f further new
Pred i c t ion of Freez i ng and Thaw i ng T i mes 1 89

terms i nto Eqs . ( 9 . 1 1 ) and/ or ( 9 . 1 2 ) ( or deve l opment of a new emp i r i cal


formula ) , to relate EHTD and MC P more c lose ly to a 1 , a 2 and B i wou ld b e
d i f f i cu l t to j u s t i fy . I t must be rememb ered that the S t e and Pk
e ff ec ts ( of up to ±5% ) have b een i gnored .

Equa t i ons ( 9 . 1 1 ) and ( 9 . 1 2 ) were f i tted for B i i n the range 0.01 to


1 00 . 0 and a from 1 . 0 to 1 0 . 0 for f i n i t e cyl i nders and i n f i n i t e rod s ,
and a 1 and a 2 from 1 . 0 to 4 . 0 for rec tangular b r i ck s . The range for
the b r i c k shap e may appear more restr i c t i ve . However , the relat ionsh i p
o f EHTD and MC P for rec tangular br i ck s was found t o b e a summat ion of
the inf i n i te rod and the f i n i te cy l i nder terms . There for e the f i n i te
cy l i nder and i n f i n i te rod data prov ided i n format ion that all owed an
accurate form of the re l a t ionsh i p for the rectangular b r i c k shape to be
found that was ap p l i cab l e bey ond the range of the data for the
rec tangular br ick shape i tse lf .

9 .5 TEST I NG Or IM PROVED GEOM ETR IC FACTORS AGAI NST EXPERIMENTAL D ATA


FOR MU LT I-D IMENSIONAL R EGU LAR SHAPES

By use of Eqs . ( 9 . 5 ) and ( 9 . 7 ) , b oth EHTD and MC P can b e app l i ed to any


slab pred ic t ion me thod . A number of accurate s i mple pred i c t ion
formulae for both freez i ng and thaw i ng are ava i l ab l e or were developed
in Chap . 7 . A l l of these me thods are of s im i lar pred i c t ion ac curacy
and calculate process t i mes h ighly corre lated w i th numer i ca l method
pred i c t ions . Rather than t est the geometr i c factors wi th each s l ab
pred i c t ion method i nd i v idua lly , it was dec i ded to i n i t i a l ly use only
one s l ab pred i c t ion me thod . The f in i te d i fferenc e method was chosen as
i t best allows the geometry e ffect to b e stud i ed on i ts own .

The p rev i ou s , and the i mp roved ver s ions o f the expr ess ions to calculate
MC P and EHTD were tested aga i nst the Ty lose rectangular b r i ck fr eez i ng
and tha w i ng data ( Cl e l and & Ear l e 1 97 9 b and Tab l e 6 . 5 ) . A summary o f
the p red i c t i o ns i s given i n Tabl e 9 . 4 .

For Ty lose rectangular b r i ck freez i ng the mean pred i c t ion error was
off-set from zero for both . the geometr i c fac tors . The full f i n i t e
d i ffer ence me thod resu l ts a l so d i sp l ayed an off-set mean pred i c t i on
error of the same magni t ude . Cl e land et a l ( 1 982 ) showed that th i s was
Pred ic t ion of Freez i ng and Thawing T i mes

probab ly due to non- equ il i b r i um freez ing rate effects not taken into
account i n calculat ions by the f in i te d i fference me thod . U s i ng the
f i n i t e d i f ference method as the slab pred i c t ion method wou ld there fore
b e expec t ed to l ead to a s im i lar off-set of the mean.

For freez i ng , the standard dev iat ion of pred i c t i ons was low , and the
corre lat ion coe f f i c i ent ( compared wi th the full f i n i te d i f ference
method results ) was h igh for both geometr i c fac tors . Th i s ver i f i es the
assump t i on that the Ste and Pk effec ts on t slab / t and hence EHTD and
MC P wer e suff i c ient ly sma ll to be ignored , and also prov i des
con f i rmat ion that for freez i ng the lack of fit of Eqs . ( 9 . 1 1 ) and
( 9 . 1 2 ) has not decreased the overa ll ag reement with exper i mental data .

For Tylose rec tangu l ar b r i ck thawi ng , the mean pred i c t ion errors for
Eqs . ( 9 . 1 1 ) and (9.12) were not s i g n i f i cantly d i f ferent from z er o .
However , the standard dev i a t i on o f the pred i c t ion error was h i gher than
that for the full finite d i f ference method pred i c t i ons and the
corre l a t i on coe f f i c i ent compared wi th the fu l l f i n i te d i ff erenc e method
results was low , for both the geometr i c factors . Th i s suggests that
the i naccuracy a r i s i ng from the improved EHTD and MC P formulae is
suff i c iently large to be d i s t ingu ished from the exper imen tal
uncerta i n ty . Two reasons ex i s t . F i rstly the rec tangular b r i ck tha w i ng
exper iments had less exper i mental uncer ta inty than the rec tangular
br i ck freez i ng data so the lack of f i t is more d i sc er n i b l e for thawi ng
but not for freez i ng . Secondly , i t was shown in Sec . 9 . 3 that the
e ffec t of Ste and Pk on t slab / t was a s much as ±5% dur i ng thaw i ng .
The se effects were ignored by Eqs . ( 9 . 1 1 ) and ( 9 . 1 2 ) , so correlat ion o f
p r ed i c t ion error s wi th S t e o r P k would ind i cate whe ther th i s was a
contr i b ut i ng fac tor . The correlat ion coe ff i c ients w i th Ste and Pk were
- 0 . 73 and 0 . 06 r espec t i ve l y , i nd i cat i ng that neg l ec t of the e f f ec t of
Ste was s i gn i f i cant .

Al though better pred i c t i ons for thaw i ng cou ld be ach ieved by add ing
terms i nvo lv i ng Ste, espec ial l y , i n to Eqs. ( 9 . 1 1 ) and ( 9 . 1 2 ) th i s s t ep
was not taken as the improvement would on ly be ev ident w i th very
accurate exper i mental data . In commerc i a l prac t i c e , food e ng i neers
mak i ng thawi ng t ime p red i c t i ons cannot determine data to represent the
cond i t i o ns as accurate l y as i s pos s i b l e in a researc h l aboratory , so
Pred i c t ion o f Freez ing and Thaw i ng T i me s 191

the ex tra t e rms wou l d not b e he lpfu l t o them .

An interest ing facet of Tab l e 9 . 4 i s that ov era l l the o ld MC P and EHTD


formu l ae perform as wel l as the new on es . Th i s might sug ges t that the
introduc t i o n of the new formulae was not worthwh i l e . Howev e r , i t must
be noted that the exper imental data se t does not cover as wi de a range
of cond i tions or geome tr i es as the numer i ca l data set of Tab l e 9.3 .
The o l d EHTD and MC P formulae are inaccurate i n parts of the w i der
range o f cond it ions and shapes covered by Tab l e 9 . 3 so the ir u se for
some B and Bi comb inat ion s , or shapes other than rectangular b r i ck s ,
cou l d l ead to subs tan t i a l pred i c t ion error s . There fore the i r
rep lacement by Eqs . ( 9 . 1 1 ) and (9.12) is j u st i f ied because o f the
reduced p red i c t i on inaccuracy r i sk s , desp i t e some add i t ional
comp l ex i t y .

9.6 TEST I NG OF IM PRO VED GEOMETR IC FAC TOR S I N COMBI NAT ION WITH
S IM PLE PRED ICTION FORMU LAE

The pred i c t ions of the imp roved geometr i c fac tors in comb ina t ion w i th
the s i mp l e fr eez i ng and thawing t i me pred i c t ion me thods of Sec . 7 . 3 are
summar i sed i n Tab l e 9 . 5 .

For free z i ng they a l l p er form simi l ar l y - those app ear i ng to be bet t er


due to l ower standard dev i at ions also have l ower corre l at i on
coeff i c ients w i t h the full f i n i t e d i fference method pred ict i ons . The
means are not off-set from z ero l ike the f i n i te d i f f erence mean , but
the standard dev iat i ons are o f s i m i l ar size to the finite d if f er e nce
me thod s tandard dev i a t ion . The off- se t mean for the f in i te d i f ference
method pred i c t ions ar i se s from the rate e ffec ts d i scussed b y C l e l and et
a l ( 1 9 82 ) . The c loseness of the standard dev i a t ion for the finite
d if ferenc e me thod and those for the s i mp l e methods ind icates that for
freez i ng t i me pred i c t i on any of the e i ght comb inat ions in Tab l e 9 . 5 can
b e used w i thou t i ntroduc ing s i gn i f i cant pred i c t ion method uncerta inty .

The s i tuat ion for thawi ng t i me pred i c t ions was s im i l ar . All e i ght
method comb i na t i ons are comparab l e , but the neg lect of Ste and P k
e ffec t s o n EHTD and MCP has b o t h l owered the corr e l a t ion coeff i c i ent
with the f i n i t e d ifference method pred i c t ions , and l ed to the standard
Pred i c t ion of Freez i ng and Tha w i ng T i mes 1 92

dev iat ion b e ing s l ight ly h i gher than for the f i n i t e d i ff erence metho d .

Because E q . ( 9 . 1 1 ) and ( 9 . 1 2 ) were de ve loped from numer ical data they


are app l i cab l e over a wide range of cond i tion s . For a l l the regular
shap es cons i dered the l im i ts are :
0.0 < Bi < 1 00

1 .0 < 8 1 < 10

1 .0 < 82 < 10

The formulae may b e app l i cab l e outs i de these ranges but at the expense
of reduced ac curacy . Howe ver , v irtua l l y a l l p rac t ical food fr eezing
and tha w i ng s i tua t ions are covered .

9 .7 SUMM ARY

The e ffec t of geometry on free z i ng and thaw i ng t i mes , ev en for regular


mu l ti -d i mens ional shap es , is suff ic iently comp l ex that no accurate
geometr i c fac tor has been der i ved ana ly t i ca l l y .

There are two useful conc ep ts to accoun t for geome try - EHTD and MC P .
Formulae to calculate both geometr i c factors for fr eez i ng of
rec tangular b r i c k shap es under a l i mi ted range of cond i t ions had been
prev iously dev e loped by emp ir i cally f i t t i ng rectangular bri ck freez ing
exper imental dat a .

B y f i tt ing a n a lt erna t i v e data set calculated by numer i ca l methods ,


i mproved formulae for MC P and EHTD were de ve l oped . These are accurate
for both freez ing and thaw i ng of an ex tended range of regular
mul t i - d i mens ional shap e s , and for a wi der range of cond i t ions than the
or i g ina l formulae .

Used in comb ina t ion w i th accurate slab fr eez ing and thaw i ng t i me
pred i c t ion me thods , b o th of the i mproved geome tr i c fac tors gave good
f i t to exper imental data for freez i ng and thaw i ng of rec tangular
b r i ck s . N e i ther of the geome t r i c factors introduc ed s i gnif i cant
p red i c t ion uncerta inty .

For p re d i c t ion of freez i ng and thawi ng · of regular shaped


mul t i -d i mens ional obj ec ts there were no grounds to re commend e i th er o f
Pred i c t ion of Freez i ng and Thawi ng Times 1 93

the two geometr i c conc ep ts ahead of the other ( s im i lar ly , none of the
slab pred i c t ion approaches stood ou t as b e ing s i gn i f i cantly more
accurate or s i mp l e to u se ) .

Tab l e 9.1 The Effec t of B i , Ste and Pk on the Rat io of Free z i ng and
Thaw i ng T i mes For I n f i n i t e Rods To the Times For the
Equ i va l ent Slab

B1 Bi Ta T in t slab /t Effect of

1 .0 0.1 - 40 . 0 20 . 0 1 . 99
1 .0 0.1 - 30 . 0 20.0 1 . 99 Ste
1 .0 0. 1 -20 . 0 20 . 0 2 . 00

1 .0 1 .0 -40 . 0 5.0 1 . 85
1 .0 1 .0 -30 . 0 20 . 0 1 .91 Ste and Pk
1 .0 1 .0 - 20 . 0 35 . 0 1 . 96
1 .0 1 .0 5.0 -1 0 . 0 1 .9
1 .0 1 .0 20 . 0 -20 . 0 1 . 8� Thaw i ng
1 .0 1 .0 35 . 0 -30 . 0 1 . 80

1 .0 1 0.0 -30 . 0 5.0 1 . 69


1 .0 10.0 -30 . 0 20 . 0 1 .72 Pk
1 .0 1 0.0 -30 . 0 35.0 1 . 75

2.0 0.1 - 40 . 0 20 . 0 1 . 47
2.0 0.1 - 30 . 0 20 . 0 1 . 48 Ste
2 .0 0. 1 - 20 . 0 20 . 0 1 . 48

2.0 1 .0 - 40 . 0 5.0 1 . 26
2.0 1 .0 -30 . 0 20.0 1 .31 Ste and Pk
2.0 1 .0 -20 . 0 35 . 0 1 . 35

2.0 1 .0 5.0 -1 0.0 1 . 35


2.0 1 .0 20 . 0 -20 . 0 1 . 26 Thawing
2.0 1 .0 35 . 0 -30 .0 1 . 23

2.0 10.0 -30 . 0 5.0 1 .10


2 .0 10.0 -30 . 0 20 . 0 1 .1 1 Pk
2.0 10.0 -30 . 0 35 . 0 1 .12
Pre d i c t ion of Free z i ng and Thawi ng Ti mes 1 94

Tab l e 9 . 2 Constan ts For Pred i c t ion Of EHTD and MC P

Shap e Gl G2 G3

S l ab s 0 0
I n f i n i te C y l i nder s 2 0 0
Spheres 3 0 0
F i n i te Cy l i nder s 2 0
o r ;:: D Y
F i n i t e C y l inders 2 0
or s Dy
I n f i n i t e Rods 0
Two- D i mens iona l 0
Irregular Shap es
Rec t angular Br i ck s
Three- D i mens iona l
Irregular Shapes
Pred i c t ion of Freez ing and Thaw ing T i mes 1 95

Tab le 9 . 3 Summary of Percentage D i f ferences Between Numer ical ly


Calcul ated Freez i ng T i mes For F i n i te Cyl inder s , Inf in i te Rods
and Rec tangular Br i cks and Freez i ng T i mes Calculated By
S imp le Pred i c t ion Formu lae 1

Shap e Geome t r i c Equat ion Mean Std Oev 2 Min 2 Max 2


Fac tor (%) (%) (%) (%)

EHTD (9.3) 3.8 7.2 -7 . 0 24 . 8


All Data EHTD (9 .1 1 ) -0 . 5 1 .6 -7 . 0 5.2
( 270 runs ) MC P (9.4) -2 . 4 9.2 -31 .5 25 . 0
MC P (9.1 2) -0 . 3 1 .7 -7 . 0 4. 1

F i n i t e C y l inders EHTD ( 9.3 ) 0.1 3.9 -6 . 6 7.8


Or � Dy EHTD (9.1 1 ) -0 . 4 0 .8 -2 . 5 1 .2
( 63 runs ) MC P (9.4) -1 1 . 3 10.5 -31 . 5 0.3
MC P (9.12) 0 .5 1 .3 -2 . 2 4.1

F i n i te C y l i nders EHTD (9.3) 6.4 9.0 -5 . 9 24 . 8


Or � Dy EHTD (9 . 1 1 ) -0 . 5 2.3 -7 . 0 5.0
( 5 4 runs ) MC P (9 . 4) -4 . 2 6.3 - 25 . 7 0.3
MC P (9.12) -0 . 7 2. 1 -7 . 0 3.5

EHTD (9.3) 3 .6 4.7 -2 . 9 1 3.8


I n f i n i t e Rods EHTD (9 . 1 1 ) -0 . 6 1 .2 -3 . 7 2.4
( 6 3 runs ) MC P (9.4) 4.2 7.1 -1 .2 25 . 0
MC P ( 9 . 1 2) -0 . 5 1 .3 -4 .2 2. 1

EHTD (9 . 3 ) 4 .9 8.2 -7 . 0 24 . 4
Rectangu l ar
Br icks EHTD (9.1 1 ) -0 . 5 1 .7 -4 . 4 5.3
( 90 runs ) MC P (9.4) 0.3 4.7 -7 . 6 1 2.5
MC P (9. 1 2) -0 . 7 1 .7 -6 . 0 3. 1

1 U s i ng the f i n i t e d if ference method as the slab pred ict ion method .


2 S td Dev s tandard de viat ion , M i n
= m i n imum, Max
= max i mum. =
Pred i c t ion of Freez i ng and Thawing T i mes 1 96

Tab le 9 . 4 Summary o f Percentage D i f ferences Between Exper imental


Freez i ng and Thawing Times For Tylose Rectangular Br icks and
Freez ing and Thawing Times Calculated Us ing S i mple Geometr i c
Fac tors 1

Data Geometr i c Equat ion Mean Std Dev 2 Min 2 Max 2 Corr 3
Fac tor (%) (%) ( %) (%) FDM

FDM .. -4 . 4 6.6 - 20 . 7 9.8 1 . 00


Freezing FEM s .. -5 . 9 7.1 -24 . 1 9 .8 0 . 97
( C l e land & EHTD (9 . 3 ) -5 . 7 7.3 -23 . 0 9 .8 0 . 90
Ear l e 1 979b )
EHTD (9 . 1 1 ) -3 . 8 6.6 -21 3 • 9 .2 0 . 86
( 72 runs )
MC P (9 .4) -6 . 0 5.5 -1 6.8 7.3 0 . 67
MC P (9 . 1 2) -4 . 9 7. 1 -22 . 7 10.9 0 . 89

FDM .. 0.6 3. 1 -6 . 0 7.9 1 . 00


FEM s .. -6 . 3 3.4 -1 2 . 9 1 .2 0 . 66
Thaw i ng
EHTD (9.3) -3 . 5 5.4 -15.0 6.4 0.06
( Tab le 6 . 6 )
EHTD (9 . 1 1 ) -2 . 2 5.3 -1 2 . 9 10.9 0 . 06
( 68 runs )
MC P (9 .4 ) -0 . 6 5.8 -1 2 . 6 13.7 -0 . 0 3
MC P (9 . 1 2) -3 . 4 5.2 -1 3.0 7.1 0. 1 1

1 U s i ng the f i n ite d if ference method as the slab pred ict ion method .
2 Std Dev s tandard dev iat ion , Min min imum, Max
= = max imum. =

3 Correlat ion coe f f ic ient ( r ) compared with the percentage d i ff erences


for the full f in i te d i f ferenc e me thod •

.. FDM = full f in i t e d ifference method , FEM s s impl i f ied f i ni t e e l ement


me thod .
Pred ict ion of Freez ing and Thawi ng Times 1 97

Tab l e 9 . 5 Summary of Percentage D i fferences Between E xper imenta l


Freez i ng and Thawi ng T imes For Ty l ose Rectangular Br i ck s and
Freez i ng and Thaw i ng Times Calculated By S imple Pred i c t i on
Formu l ae

Data Slab Pred ict ion Geometr ic 1 Mean Std Dev 2 Min 2 Max 2 Corr 3
Me thod Factor (%) (%) (%) (%) FDM

FDM " -4 . 4 6.6 - 20 . 7 9.8 1 . 00


EHTD -0 . 1 4.9 -8 . 9 9 .7 0 . 57
Eq. ( 7 . 7 )
MC P -1 . 3 5.2 -1 1 . 0 8.0 0 . 66
Freez i ng
Cl e land & EHTD 0.8 5.3 - 9 .0 15.5 0 . 49
( C le l and & Ear le 1 9 82b
Ear le 1 97 9 b ) MC P -0 . 4 5.5 -1 0.5 12.7 0 .59
( 7 2 runs) EHTD -0 . 8 6.3 -1 6 . 4 1 1 .2 0 . 91
Pham 1 98 4 c
MC P -1 . 9 6.8 -17 .9 13.4 0 . 94
Pham 1 9 83 EHTD -2. 1 6.0 -1 5 . 9 1 1 .2 0 . 87
( Eq . ( 7 . 6 ) )
MC P -3 . 3 6.5 -17 .4 10.8 0 . 89

FDM " 0.6 3. 1 -6 . 0 7.9 1 . 00


EHTD -1 .6 4.3 -9 . 1 1 1 .5 0 . 28
Eq . ( 7 . 2 )
MC P -2 . 8 4.2 -9 . 8 7.7 0 . 36
Tha �o� i ng EHTD -2.3 5.3 -1 2.9 1 0 .6 0 . 36
Eq . ( 7 . 3 )
( Tab le 6 . 5 ) MC P -3 . 5 5.1 -1 3 . 4 6.6 0 . 42
( 68 runs ) EHTD -2 .6 6.0 -1 5 .6 1 1 .9 0 . 35
Eq . ( 7 . 4 )
MC P -3 . 9 5.5 -16 . 1 7.8 0 . 43
EHTD -1 . 1 4.3 -1 1 5
• 7.9 0 . 35
Eq. ( 7 .5 )
MC P -2 . 4 4.4 -1 1 .5 6.3 0 . 40

1 EHTD from Eq . ( 9 . 1 1 ) , MC P from Eq . ( 9 . 1 2 ) .


2 Std Dev standard dev iat ion , Min
= min imum, Max
= max imum. =

3 Correlation coeff ic ient ( r ) compared w i th the percent age d ifferences


for the ful l f in i te d i fference method .
" full f i n ite d i fferenc e method .
1 98

10 PREDICTION OF FREEZING AND THAW ING T IMES FOR MULT I-DIMENSIONAL


I RR EGU LAR SHAPES BY SIMPLE METHODS

T he prev ious three chap ters have covered the th ird and four th research
obj ect i ves set in Chap . 3 - assessment both of the accuracy of Group I
( s imp l e ) and Group I I ( numer i cal ) pred i c tion me thods for r egular shapes
and of Group II methods for irregular shapes . Chap ter 1 0 is concerned
wi th the f i fth obj ect i ve inves t i gat ion of the poss ib i l i ty o f
deve lop ing a Group I method t o pred ict freezing and thaw i ng t i mes for
irregular shapes .

10.1 I NTRODUCTION

No simp le methods suggested for ca lcu lat ion o f geometr i c fac tors from
f i rst p r i nc i p l es for freez i ng and thawing of irregular shaped obj ec ts
have proved accurat e . Rutov ( 1 936 ) and Mat t ( 1 964 ) used the AD/ 2V
ratio as a shape factor but it is known that th is i s not accurate when
Bi � . Smi th et al ( 1 967 ) , Smith et al ( 1 968 ) , Cl ary et al ( 1 968 ) ,
Smi th & Nelson ( 1 969 ) and Clary et al ( 1 97 1 ) def i ned a geometry i ndex
( the forerunner to EHTD ) , for the case of heat conduc t ion wi thout
change of phase and B i ��. For irregular geometr i es they calculated
the i r geometry i ndex from an e l l ipso ida l model shape that has equa l
orthogonal cross-sect ional areas and the same c haracter i s t i c th ickness
as those of the anoma l ous shape that it r eplac es .
The orthogonal
cross-secti ons taken were generally the smal l and largest that pass
through the thermodynami c c entre and l ie in the same p lane as the
character i st i c d imens ion . I t was unknown how th is method wou ld perform
for phase change and when B i -�.

For regular mul t i-d imens ional shapes use of the concepts of EHTD and
MCP led to accurate me thods to take accoun t of geometry in freez i ng and
thawing t i me pred i c tions . EHTD and MC P are calculated as funct ions o f
the Blot number and the two other geome try parameters C e 1 and e 2 ) that
d es cr ibe the regular geometr i es . It was cons i dered sens ib l e to try to
extend the methods to calculate EHTD and MC P t o irregular geometr i es .
I t was dec i ded to ignore any vari ation o f these geome t r i c factors w i th
Ste and Pk as th i s var i at ion would probably be small c ompared w i th the
Pred i c t ion o f Freez ing and Thawi ng Ti me s 1 99

B i and shape effec ts ( Sec . 9 . 3 ) .

10.2 GEOMETRY PARAMETERS

U s i ng Eqs . ( 9 . 2 ) and ( 9 . 9 ) the va lues of EHTD and MC P can be defined


from i nd i v idual exper i ments for a par t i cular shape ( Cleland & Earl e
1 98 2 b ) . Howe ver these determinat ions are both geometry and cond it ion
spec i f i c and do not nec essar i ly allow accurate and s impl e calcul at i ons
for other cond it ions or geometr ies . A major problem w i th deve lop ing
me thods to de term i ne EHTD and MCP from f irst princ ipl es is that apart
from D , A and V , there are few other eas i l y measured parameters that
can be used to charac ter i se all irregular shapes .

Even the character i s t i c ( fi r s t ) d imens ion , D, is not alway s


s t ra i g h t f o r w a r d to es timate . For an irregul ar shape D i s de fined
e i ther as ( a ) the smallest th ick ness measured by a l i ne that passes
through the thermodynam ic centr e , or ( b ) twice the shortest d i s tanc e
from the thermod ynamic centre t o the obj ects surface . The pos i t ion o f
the thermodynamic centre i s not always obv ious from examination o f the
obj ect and may mov e dur i ng the phase change process ( Fl eming 1 970 ) .
For most common shapes where surface boundary cond i t ions a re not
pos i t ion var i able ( such as those c o n s i d e r e d in th i s wo r k ) , the two
def in i t ions of D are v irt ua l l y equ i valent .

Poss ibly use ful measures of the effect of the second and th ird
d i mens ions are :
( a) The rat i o between the max imum d istance from the thermodynamic
centre to the surface and D/ 2 .
( b ) The rat io o f the major ax i s leng th to the minor ax i s l eng th for an
e l l i p t i ca l model cross-sect ions equal in area to the orthogonal
cross-sect ions o f the anoma lous shape ( Smith et a l 1 967 ) .
( c ) The ratio of the volume o f the obj ect compared w i th the volume for
a i n f in i te cyl inder ( two-d imens ional shapes ) or sphere
( three-d i mens ional shapes ) o f the same surface area .
These three parame t er s are h ighly correlated with each other as they
all measure in some way , the d i stor t ion of the shape from a inf i n i te
cylinder or sphere . Each has some d ifferent propert i es so all three
were inves ti gated to see if they could ind i v i dual l y , or i n comb ination ,
Pred ict ion of Freezing and Thaw ing T imes 200

prov i de s imple me thods to calculate geomeric factors for irregular


shapes .

A fourth but , very d ifferent , way to descr i b e an objects shape is to


def ine a regular shape that freezes or thaws i n the same t i me as the
obj ect under the same env ironmenta l cond itions . For example , Pham
( 1 98 4 c ) def ined an equ ivalent rectangular b r i ck shape for fini te
cyl inders , but th is led to uncerta inty in the shape f actor and
consequently underpred i c t ion of freez i ng and thaw i ng t i mes , espec ial ly
at high B i values ( Tab le 9 . 3 ) . Th is was because the s i z e o f the
equ ivalent b r i ck ( br ick wi th the same freez i ng or thawing t ime ) changes
w i th Biot numb e r . Al though th is approach has proved suc c essful for
some spec i f ical ly stud ied irregular shap es by approx imat ion to slab s ,
infinite cyl inders or spheres and defin ing equ iva lent d iamet ers
( Sec . 2 . 5 , Tab le A . 8 ) , it is unl ikely to lead to a me thod that can be
appl ied sy stemat ica l ly to a wide range of shapes with accuracy . I t was
dec i ded not to cons ider i t further un l ess a l l of the o ther approaches
proved unsu itab l e .

10.3 DEV ELO PMENT OF GEOMETRIC FACTOR S FOR MU LTI-DIMEN SIONAL


IRR EGU LAR SHAPES

Both EHTD and MC P are defined for any shape as B i � by the AD /2V rat io .
There fore the requ irement i s to develop a me thod to calculate the
change in these factors w i th B iot number , in as s impl e and as accurate
a manner as poss ib l e . The calculat ion me thod should only u se eas i l y
measured o r est imated geometry parameters such as those d iscussed in
Sec . 1 0 . 2 .

There are few experiment al data for free z i ng and thawing o f irregular
shaped obj ec ts ava i l ab l e in the l i terature , for wh ich the geometry and
cond it ions are suff i c iently accurate ly descr i bed , measured and
contro l l ed to a l l ow the ir u se in the pr esent s tudy . For thi s reason ,
the method of analysi s app l i ed was the same as that used for r egul ar
shap es ( Sec . 9 . 4 ) . I n i t ia l l y only the ratios of numerically calculated
freez ing and thawi ng t imes for the actual shape , and for the r eference
s l ab were used in the development of the geome tr i c factor s . Onl y the
var i at ions due to geometry and the e ffect of Bi were apparent in these
Pred i c t ion o f Freez ing and Thaw i ng T i me s 20 1

d a t a a s m i d - range S t e and Pk v a l u es were used . The cho i c e o f app roach

was e s p e c i a l ly important for the three-d imens iona l i r r eg u l ar shap es

wher e there was the poss ib i l i t y o f s y s t e ma t i c sour c e s of e x p er i me nta l

error i n the e x p er i menta l data ( Sec s . 5 . 9 . 3 , 5 . 9 . 4 and 8 . 4 ) .

E v e n though an i n f i n i te numb er of shapes are pos s i b l e , d ue to

l i m i ta t i ons in grid p repar a t ion t i me and compu ta t i on resourc e s ,

numer i ca l data were only c a l c u l ated for the e ight two-d i mens i ona l

obj e c t s and the four three-d imen s i o na l objects u sed i n exper i me n t s

( F ig s . 5 . 1 5 to 5 . 2 1 and 5 . 2 4 to 5.2 ) . A l so comp u t e r proc ess t i me


7
l i m i ta t i ons meant that computa t ions were not pos s i b l e for the f ul l

r ang e o f Bi values ; only the impor t a n t case wh e r e B i +� was c a l c u l a t ed .

The data are g i ven i n Tab le C . 2 . To suppl ement the data base , the

numer i ca l da t a for the regu l ar shap es were a ls o used ( Tab l e C . 1 ) . Any

me thod to ca l c u l a t e the geome t r i c factor for i r regu l a r shap es shou l d

a l s o b e ap p l i cab l e to reg u l ar shap es . Howe ver , lower accuracy compared

wi th the me thod s s p ec i f i c for regu l a r shap es , descr ibed i n Chap . 9 ,

wou ld be e x pec ted b ecause for the method to be u n i v e r sa l , less deta i l ed

i n format ion about the geome t r y can be i ncorporated . The r egu l ar shape

data both i ncreased the range o f geome t r y t yp e s and added i n format ion

abou t how the geome t r i c factors behave at i n t e r med i ate B i v a l u es . The

compos i te data set was s t i l l l i m i t ed in scope .

B y c urve f i t t i ng of th e numer i ca l data the fo l l ow i ng e xp r es s ions were

found to be of reasonab l e accurac y , b u t as s i m p l e as pos s i b l e :

EHTD G1 + G2 [� - ]( 1 f 1 5( V 1 ) + (1 - f l 5( V l ) ) IV
1 .4
l 7

[ ](
=

2A 1

L ,D f l 5( V2 ) 2. 5
1
_

+ G3 + ( 1 - f l 5( V2 ) ) I V2 7 ( 10. 1 )
2A 2

(
and
MC P
1 + ( 2 -G 1 ) f1 1 f 1 6 ( G2 , V 1 ) ) 0 . 65 B i 0 · 39 ( V 1 · 39 - 1 )

(
=

+ G3 f l 1 ( f l , ( G3 , V 2 ) ) 0 . 65 B i 0 . 39 v; · 2 6 - 1

+
rAD
L2v
I ( G1 + G2 rL l D -
I_2A 1
]
1 1 vl · 4 7 + G3 rL 0
L2A 2
2
- J
1 i v� · 7 5 ) - 1 ] o

o
(1 - f1 1 ( ( f 1 , ( G2 , V 1 ) + f 1 , ( G3 , V 2 ) ) I ( G2 G3 + 1) )) (1 0 •2)

where f 1 5 ( V ) • 3 . 56 1 v1 • 77 I ( Bi 1 · 3 4 + 3 . 56 1 v 1 · 7 7 )
f 1 ,( G, V) • G 3 .08/ v •
1 90
Pred iction o f Freezing and Thawing T imes 202

• X I ( B i 1 · 35 + X )
z per imeter l eng th o f the first and second
orthogonal cross-sec t ions (m)
area of the f ir s t and second or thogona l
cross-sect ions
rat io o f major a x i s leng th to minor ax i s l eng th for
an e l l i p se mode l l i ng an or thogona l cross- sec t ion
4A I nD 2
1

4A 2 1 nO 2
and the va lues of G1 , G2 and G3 are g i ven i n Tabl e 9 . 2 . Tab l e 1 0 . 1
g i ves the geome try parame ters used in Eqs . ( 1 0 . 1 ) and ( 1 0 . 2 ) for the
shapes used in th is study ( Tab les 6 . 6 and 6 . 7 ) .

Th ese equat ions use the second o f the shape parameters d i scussed in
Sec . 1 0 . 2 y e t are st i l l defin ite extens ions o f the methods deve loped i n
Sec . 9 . 4 f o r regular shap es . The techniques are s imi lar to that o f
Smith et al ( 1 967 ) in that the or thogona l cross - sect ions that pass
through the thermodynamic centre are used and the contribut ions to the
geometr i c factor calculated from each cross- sec t ion are summed ( th i s i s
also the case i n Eqs . ( 9 . 1 1 ) and ( 9 . 1 2 ) ) . Due to the l ack o f data ,
accurate f i t t i ng of all the coef f ic ients was not poss ible so many of
the coe f f i c ients have b een defined by ana logy to the methods developed
for the regular shape geometric factors ( for example ,
3 . 081 v 1 · 7 7 approx i mates 2 . 3 21 a 1 · 77 and B i 1 · 3 4 is reta ined to g ive the
B iot number we igh t i ng func t ions in the equat ion for EHTD ) . Equa t i ons
( 1 0 . 1 ) and ( 1 0 . 2 ) are very s i m i lar to Eqs . ( 9 . 1 1 ) and ( 9 . 1 2 ) because V

i s a parameter tha t i s c lose ly re lated to a for r egular shap es .


However , th i s means that for shapes increas ing l y d i storted from r egul ar
shapes the accur acy may be reduced . All of the irregular shapes used
in the ana lys i s were e i ther relative ly oval w i th no shar p contour
changes or were rectangular with square corners or edges ; so th i s was
not a probl em wi th th i s data set .

Tabl e 1 0 . 2 shows the f i t o f Eqs . ( 1 0 . 1 ) and ( 1 0 . 2 ) to the numer i ca l


data for regu lar shapes . Overal l the agreement i s good and i s only
sl ightly inferior to that d isplayed by the spec ial i sed methods ( Table
9.3) .
Pred ict i on o f Freez ing and Thawi ng T i me s 203

10.4 TESTI NG OF GEOMETRIC FACTORS AGAI NST EXPERIMENTAL DATA FOR


MU LT I- DIMENSIONAL SHAPES

Equat ions ( 1 0 . 1 ) and ( 1 0 . 2 ) can be used in conj unct ion w i th any of the
slab pred i c t ion me thod that had prev iously been used w i th Eqs . ( 9 . 5 )
and ( 9 . 7 ) , to pred i c t freez ing and thawi ng t imes for objects o f any
shape . Table 1 0 . 3 summar i ses the pred i c t ions aga i ns t the exper imental
data of C l eland & Ear le ( 1 979b) and Tab les 6 . 5 , 6 . 6 and 6 . 7 for
rec tangular b r i ck s , two- d i mens ional irregular shap es and
three-d imens iona l irregular shapes .

1 0. 4 . 1 Rec tangular Br ick Freez ing and Thawing

The pred ic t ions were almost iden t i cal to those d i splayed for E q . ( 9 . 1 1 )
and ( 9 . 1 2 ) in Tab les 9 . 4 and 9 . 5 . Th is con firms that for regular
shapes , the general calculat ion me thods for EHTD and MCP ( Eq s . ( 1 0 . 1 )
and ( 1 0 . 2 ) ) are equ ivalent to the spec ial i se d methods ( Eq s . ( 9 . 1 1 ) and
( 9 . 1 2 ) ) , and do not add s i gn i f icant unc erta inty to freez ing or thaw i ng
t ime pred ict ions .

10.4.2 Freez ing and Thaw ing of Two-D imensional Irregular Shapes

For the two-d imensiona l irregular shape dat a the mean pre d i c t ion errors
us i ng E q . ( 1 0 . 1 ) and ( 1 0 . 2 ) to descr ibe geometry were sl ightly off- set
from z ero for both the geometr i c factors , and for both freez i ng and
thaw i ng . The standard dev iat ion was low and the correl a t i on
coe f f i c ient ( compared with the full f in it e e l emen t method results ) was
h igh . The magni tude of the 9 5 % conf i dence bounds for the pred i c t i on
errors was s imilar to the est imated exper imental error bounds for th i s
data ( Sec . 5 . 8 . 4 , Tab le 1 2 . 2 ) . Hence for the two-d i mens ional irregu l ar
shapes used ( all of wh ich had curved , ova l surfaces ) the lack of f i t o f
the geome tr i c fac tors calculated by Eqs . ( 1 0 . 1 ) and ( 1 0 . 2 ) does not
s i gn i f i cantly worsen the over a l l agreement w i th exper i ment al data
a l though there i s a sl i ght sh i ft o f the mean.
Pred ict ion o f Freez ing and Thawing Times 20 4

10.4.3 Freez i ng and Thaw i ng of Three-D imens iona l Irregular Shapes

I n general the accuracy o f pred i c t ions for the three-d i mens ional
irr egular shapes was poorer than for the other shapes . Th is was
expected due to a h i gher exper imental unc erta inty for these shapes
( Sec . 5.9.4). The st andard dev iat ion o f the pred ictions were s i m i lar
to those for the ful l f i n i te e lement method r esu lts and the corre l a t ion
coe ff i c ient of the percentage d ifferences were h igh.

However , the mean pred i c t ion errors d i splayed d ifferen t trends to those
for the full f in i t e element results ( Tabl e 8 . 2 ) . For the py rami d shape
the mean pred i c t ion error was closer to z ero compar ed wi th the
under-pred ict i on shown by the full f i n ite e lement method results . As
d i scussed in Sec . 8 . 4 the off-set for the numer ical results was partly
caused by use o f a coarse spat ial gr id for the ca lcu lat ions . The
s i mp l e slab pred i c t ion me thods and Eqs . ( 1 0 . 1 ) and ( 1 0 . 2 ) were not
affec ted by th is problem so , as expec ted , for an accurate slab
pred i c t ion me thod the mean pred i c t ion error was not s i gn i f icant l y
d i fferent from z ero .

The same explanat ion app l ies for the sphere shape results . The
pred i c t ions for the ful l f in i te e lement me thod us ing a
three-d imens ional g r i d were o ff-set from z ero , but both the pred ict ions
for the f in i te e l ement method us i ng a f iner one-d i mens iona l g r i d and
those for the s impl e predict ion methods were not .

For the egg and f i sh shapes d ifferent trends were observed . The mean
pred iction error was h igher for the s i mple pred iction methods than the
nume r i cal me thods for the egg shape , but l ower for the f i sh shape . It
was d if f icult t o assess whether the d if ferenc es i n the mean p red ict ion
error s were apparent due to systematic error in the geome t r i c factors
for these shapes , because the numer i cal ca lculat ions were i n f luenced by
factors such as the coar se spatial grid ( Sec . 8 . 4 ) , or becau se the
pred i c t ions were affec ted by systemat i c error in the value of the
surface heat transfer coeffic ient used in calculations ( Sec . 8 . 4 ) . All
three reasons probably cont r i bu ted i n b oth case s .
Pred i c t ion of Freez ing and Thawing T imes 205

Because the pred i c t ions for the two-d i mens ional irregu lar shapes were
based on data for shapes w i th both rectangular and ova l cross-sections
and were qu i te prec ise , it was cons idered that the effect of the second
d imens ion was accurately mode lled by Eqs . ( 1 0 . 1 ) and ( 1 0 . 2 ) for both
ovo id and rectangular shapes . There fore any uncerta inty d ue t o the
geometr i c factors for the egg and f i sh shapes was thought to be in the
mode l l i ng of the e ffect of the th ird d i mens ion . Thi s is not surpr i si ng
as there were very few data ava i lab le to ana ly s i s the e ffect o f the
th ird d imens ion in three-d i mens ional irregul ar shap es . The ma i n r eason
for th i s was the large computat ion costs involved in numer i ca l
calculat ion for three-d i mens iona l ly irregular shapes . A typ i cal
calculat ion using the full f i n ite e lement method for the f i sh shape
took 1 50 000 sec . proc ess t i me on a Pr i me 750 computer . Th is meant
that the ana ly s i s was forced to re ly heav i ly on the rectangular brick
and f i n i te cyl inder shape data . These regular shapes have rectangul ar
cross-sect ions in the th ird d imens ion , not ova l as found in the eg g and
f i sh shap es . Consequently the coe f f ic ien ts in Eqs . ( 1 0 . 1 ) and ( 1 0 . 2 )
f i t ted data for shapes w i th rec tangular cross-sect ions in the th ird
d i mens ion and do not seem to be as accurate for shapes wi th other
geometries . To deve lop more accurate methods to calcu late the
geome tr i c fac tor s , EHTD and MC P , more data are requ ired . I f more
three-d imens i ona l irr egular shape data were ava i l ab l e , inc lud ing data
for shapes w i th oval cross- sec t ions in the th i rd d i mens io n , i t was fe l t
that the princ iples o f the methods to calculate EHTD and MC P ( Sec . 9 . 4 )
woul d not need to be changed ; only the values o f the coe f f i c ients i n
Eq . ( 1 0 . 1 ) and ( 1 0 . 2 ) ( and poss ibly the geometry parameter u sed ) , wou l d
a l ter to f i t the data .

Overall , the off-se t o f the mean for the pred ict ions using Eq . ( 1 0 . 1 )
and ( 1 0 . 2 ) was not cons idere d large i n relation to the inaccurac y o f
the exper i menta l data and the lac k o f su i tab le data to deve lop and test
the geomet r i c factor expr ess ions for three-d i mens ional i rregular
shapes .
Pre d i c t ion o f Freez ing and Thawing Times 206

10.4.4 Compar i son o f Slab Pred i c t ion Methods

A l l o f the f i ve s l ab pred i c t ion me thods ( the f inite d i fference method


plus the four simp l e formulae approaches ) , use d in conj unct ion w i th the
expr ess ions for EHTD and MCP gave adequate pred i c t ion ac curacy . As
d i sc ussed for rec tangular brick freez ing ( Sec . 8 . 2 ) , use o f numer i ca l
slab me thods lead t o under-pred i c t ion compared wi th the other s impl e
slab pred ict ion methods ( Chap . 7 ) . Otherwi s� the mean pred ict ion error
was almost i den ti cal for all the slab pred i c t ion methods . The f i ve
slab pred ict ion methods a lso gave very s im i lar standard d e v iat ions o f
pred ic tion error s and corre lation coeff ic ients w i th nume r i ca l method
pred ict ions .

10.4.5 Ana ly s i s o f Geometr i c Fac tors

T he accuracy of the p red ictions ob ta ined using the simp l e geometr i c


factors , EHTD an d MC P , was suffic iently good for mo s t of the shapes
tested that no further geometry parameters or geometr ic fac tor concepts
were i nv es t igated .

The expressions for EHTD and MC P gave a lmost ident i ca l freez ing and
thawi ng t ime pred i c t ions for a l l shapes and could not be d i ff eren t iated
by the accuracy c r i ter ion alone . The express ion for EHTD was sl ightly
l ess complex i n i ts f inal form than the express ion fo r MCP . Becau se o f
the peak in MCP values at intermed iate B iot numbers it was mor e
d i f f i cu l t to f ind and curve-fi t a su i table expr ess ion fo r MCP than i t
was for EHTD .

I t was d i ff icult to assess the ranges o f app l icab i l i ty of Eqs . ( 1 0 . 1 )


and ( 1 0 . 2 ) because o f the large var iety o f geome tr i es that are common
in prac t i ce . The ir accuracy has been p roven for two-d imens ional shapes
wi th both ova l and rectangular cross- sec t ions so i t seems reasonab l e to
suggest that the equat ions should apply equally we ll to mos t
two-d i mens ional obj ect s . For three-d imens ional shap es Eqs . ( 1 0 . 1 ) and
( 1 0 . 2 ) have only been p r ov en accurate for obj ects with rectangular
cross-sec t ions in the th ird d imens ion . Therefore the l im i t s of the i r
app l i ca t ion should b e t o objects w i t h th is type o f geometry . The
formulae c an be u sed for o ther shap es , but at the expense o f reduced
Pred ict ion o f Freez i ng and Thawi ng T i me s 207

accuracy . The accuracy of Eqs . ( 1 0 . 1 ) and ( 1 0 . 2 ) for both two- and


three-d imens iona l objects with extreme d is tortion of geometry wou ld b e
expected to be lower . More ref inement of the expr ess ions for ETHD and
MC P can be j us t i f ied , espec ially to mak e them app l icab l e to a w i der
range of three-d i mens ional geometr ies .

The way in wh ich the equat ions for EHTD and MCP take account o f the
effect of Bi was thought to be un i ver sa l ly correct so the l imi ts o f
app l icab i l i t y w i th regard t o th is parameter are the same a s for
Eqs . ( 9 . 1 1 ) and ( 9 . 1 2 ) for regular shap es :
0.0 < Bi < 1 00 .

1 0 .5 SUMMARY

The princ iples under ly ing the methods to ca lcu late geometr i c fac tors
u s l ng the EHTD and MCP conc epts were sat i sfactori ly extended to
irregular geometr ies . In a simi lar manner to that used for regular
shap es in Chap . 9 and by Sm i th et al ( 1 967 ) for irregular shap es , the
effect o f each d imens ion was calculated separately by cons ider ing each
of the orthogonal cross-sec t ions that pass through the thermodynam ic
centre . The parame t er �=4A/ no 2 for these cross-sect ions was s imilar to
the parameter 8 used for regular shapes and allowed expres s ions to
ca lcul ate the two geome tr i c factors to be developed along s imi lar l i nes
to those used for regular shapes .

Used i n conj unct ion w i th accurate slab freez ing and thawi ng t ime
pred i c t ion methods the express ions developed for EHTD and MCP both gave
pred ictions of s i m i l ar accuracy to those ach ieved by numer i ca l methods
for rectangu l ar b r i cks and two-d i mens ional irregular shapes . The
geometr i c factors d id not add s i gn i f i cant uncertainty to the
pred ic t ions for these shapes .

For the three- d i mens ional irregular shapes the accuracy of the
geome tr i c factors was l ower than for the two-d i mens ional irregular
shapes . Th i s was attribu ted to l ess accurate modell ing o f the e ffect
o f the th ird d i mens ion r ather than error in mode l l i ng of the effect o f
the second d imension o r because the princ iples o f the calcul at ion
method were i napprop r i ate as the pred i c t ions for the two-d i mens i onal
Pred ict ion o f Free z i ng and Thaw i ng Time s 208

shapes were accurate .

The methods developed are accurate for most two-d imens iona l objects but
are rest r i c ted to three-d i mens ional obj ec ts w i th rectangular
cross-sections in the th ird d imens ion . For these shapes the methods do
not add s ign i f i cant ly to the pred i c t ion unc erta inty . For o ther shapes
lower accuracy was apparent .

The present ana l y s i s was l im i ted by the smal l data set , and by
d i ff icul t i es in obta i n i ng accurate exper imenta l measurements espec ially
for three-d imensiona l irregu lar shapes . More da ta wou ld enab l e more
accurate and un i ver sa l ly app l i cable express ions to ca lculate EHTD and
MC P to be deve loped . Sub stant ial comput i ng costs wou ld have to be
borne in order to do th i s wor k .
Pred ict ion of Freez ing and Thaw i ng T i me s �9

Table 1 0 . 1 Parame ters For Calculat ion of the Effec t of Geometry For
Irr egular Shap es

Shape 1 0
& Code
( m)

2 01 0 . 1 525 0 . 0 1 83 0 . 479 0 . 0 1 83 0 . 479 1 . 00


2DI 2 0 . 1 1 1 5 0.01 64 0 . 475 0 .01 64 0 . 475 1 . 68
2 01 3 0 . 1 370 0 . 0 1 7 7 0 . 4 75 0 . 0 1 77 0 . 475 1 . 20
201 4 0 . 0990 0 . 0 1 52 0 . 478 0 . 0 1 52 0 . 47 8 1 . 97
2 01 5 0 . 1 050 0 . 00866 0 . 330 0 . 00866 0 . 330 1 . 00
20 1 6 0 . 0850 0 . 00796 0 . 323 0 . 00796 0 . 323 1 . 40
2 01 7 0 . 09 1 0 0 . 008 1 1 0 . 324 0 . 0081 1 0 . 32 4 1 . 25
201 8 0 . 0 475 0 . 00586 0 . 325 0 . 00586 0 . 325 3 . 30

Pyramid 0 . 0 825 0 . 00 1 3 6 0 . 0780 0 . 0 1 03 0 . 41 6 1 . 93 0 . 0 1 1 5 0 . 460 2.15


Sphere 0 . 1 2 70 0 . 00 1 07 0 . 0507 0 . 0 1 27 0 . 399 1 . 00 0 . 0 1 27 0 . 399 1 . 00
Egg 0 . 1 700 0 . 00 3 63 0 . 1 1 75 0 . 0227 0 . 534 1 . 00 0 . 0 320 0 . 652 1 . 41
F i sh 0 . 1 200 0 . 00 4 3 7 0 . 1 625 0 . 01 43 0 . 424 1 . 26 0 . 0 4 88 1 . 076 4 . 31

1 20 1 two-d imens ional irregular shap e , code number from Tabl e 6 . 6 .


2 For the 2 0 1 shapes these va lues are for a 1 . 0 m long object .
Pred i c t ion of Freez i ng and Thaw i ng T i mes 210

Table 1 0 . 2 Summary o f Per centage D i f ferences Between Numer i ca l ly


Calculated Freez ing Time s For F i n i t e Cy l inder s , I n f i n i t e
Rods and Rectangular Br ick s and Freez ing Times Ca lculated By
S imple Pred ict ion Formulae 1

Shape Geomet r i c Equa t i on Mean Std Dev 2 Min 2 Max 2


Factor (%) (%) (%) (%)

EHTD (9. 1 1 ) -0 .5 1 .6 -1 . 0 5.2


All Dat a EHTD (1 0 . 1 ) -0 . 3 1 .7 -6 . 8 5.7
( 270 runs ) MC P (9.12) -0 . 3 1 .7 -1 . 0 4.1
MC P (1 0.2) -0 . 3 1 .7 -6 . 9 4.2

F i n i t e Cyl inders EHTD (9 .1 1 ) -0 . 4 0.8 -2 . 5 1 .2


Dr � D y EHTD (1 0 . 1 ) -0 . 4 0.8 -2 . 8 1 .2
( 63 run s ) MC P (9.12) 0.5 1 .3 -2 . 2 4.1
MCP ( 1 0.2) 0.2 1 .5 -4 . 1 4.2

F i n i te C y l inder s EHTD (9. 1 1 ) -0 . 5 2.3 -7 . 0 5.0


D r i:: D y EHTD (10 . 1 ) -0 . 3 2 .5 -6 . 8 5.7
( 5 4 runs ) MC P (9 . 1 2 ) -0 . 1 2.1 -1 . 0 3 .5
MCP (10.2) -0 . 5 2.2 -6 .9 4.0

EHTD (9.1 1 ) -0 . 6 1 .2 -3 . 7 2.4


I n f i n i t e Rods EHTD (1 0 . 1 ) -0 . 3 1 .3 -3 .7 2.6
( 6 3 runs ) MCP (9.12) -0 . 5 1 .3 -4 . 2 2.1
MC P ( 10.2) -0 . 3 1 .3 -4.2 1 .9

EHTD (9. 1 1 ) -0 . 5 1 .7 -4 . 4 5.3


Rectangular
Br i c k s EHTD ( 10. 1 ) -0 . 2 1 .1 -4. 3 5.6
( 90 runs ) MCP (9.12) -0 . 1 1 .7 -6.0 3. 1
MC P ( 10.2) -0 . 6 1 .7 -6 . 1 3.9

1 U s i ng the f i n i t e d if ference method as the slab pred ict ion method .


2 Std Dev =standard dev iat ion , Mi n min imum, Max= max i mum. =
Pred ict ion o f Freez ing and Thawi ng Times 21 1

T abl e 1 0 . 3 Summary of Percentage D i fferences Between Exper i mental


Freez i ng and Thaw i ng T imes For Tylose Mu l t i - D imens ional
Shapes and Freez i ng and Thaw i ng T i mes Calculated By S i mp l e
Pred i c t ion Formu lae

Dat a Slab Pred ict ion Geometr i c 1 Mean Std Dev 2 Min 2 Max 2 Corr 3
Me thod Factor ( %) (%) ( %) (%) FDM

FDM " -4. 4 6.6 - 20 . 7 9.8 1 . 00


EHTD -3 . 4 6.7 -21 1• 9 .6 0 . 86
FDM
MC P -4 . 9 7.2 -22 . 8 1 1 .3 0 . 89
Rectangu lar EHTD 0.3 5. 1 -8 . 5 10.1 0 .56
Br ick Eq . ( 7 . 7 )
MC P -1 . 3 5.4 -1 1 1
• 8.3 0 . 65
Freez i ng
Cl e land & EHTD 1 .2 5.5 -8 . 7 15.9 0 . 48
( Cle land & Earl e 1 982b
Ear le 1 97 9 b ) MC P -0 . 4 5 .7 -10.6 13.1 0 . 58
( 72 runs ) EHTD -0 . 4 6.4 -1 6 . 2 1 1 .7 0.91
Pham 1 98 4 c
MC P -2 .0 6.9 - 1 8 .0 13.9 0 . 93
Pham 1 9 83 EHTD -1 .8 6. 1 -1 5 . 7 1 1 .4 0 . 86
( Eq . ( 7 . 6 ) )
MC P -3 . 3 6.6 -1 7 .5 1 1 .3 0 . 89

FDM " 0.6 3.1 -6.0 7 .9 1 . 00


EHTD -1 .6 5.4 -1 2 . 3 1 1 .2 0 . 06
FDM
MC P -3 . 2 5.3 -1 2 . 8 6.8 0.10
Rec tangular EHTD -1 .0 4 .3 -8 . 8 1 1 .8 0 . 29
Br ick Eq . ( 7 . 2 )
MC P -2 . 6 4 .2 -9 .8 7.4 0 . 34
Thaw i ng
EHTD -1 . 7 5.3 -1 2 . 7 10.8 0 . 36
( Tabl e 6 . 5 ) Eq. ( 7 . 3 )
MC P -2 . 2 5.1 -1 3 . 3 7.0 0 . 41
( 68 runs )
EHTD -2 . 1 5.9 -1 5 . 4 12.1 0 . 36
Eq. ( 7 . 4 )
MC P -3 . 7 5.3 - 1 6.0 7.4 0 . 43
EHTD -0 .6 4.4 -1 1 1
• 8.2 0 . 35
Eq. ( 7 .5 )
MC P -2 . 2 4.6 -1 1 7
• 7.4 0 . 38

• • • cont i nued
Tab l e 1 0 . 3 cont inued • • • 212

Data Slab Pred ict ion Geometr ic 1 Mean Std Dev 2 Min 2 Max 2 Corr 3
Method Factor (%) ( %) ( %) (%) FDM

FEM' ' - 1 .0 4.5 -1 1 .4 7.6 1 . 00


EIITD 2.5 4 .6 -8 . 3 1 1 .4 0 . 85
FDM
MC P 2.8 4.6 -8 . 3 1 1 .8 0 . 84
Two-
D imens ional EHTD 3.9 5. 1 -9 . 6 1 3 .5 o . 72
Irregular Eq. ( 7 .7 )
MC P 4. 1 5.1 -8 . 9 1 4.0 o . 72
Freez ing
Cl e land & EHTD 4.5 5.6 -9 .6 1 6.8 0 .83
( Tab le 6 . 6 ) Ear le 1 982 b
MC P 4.7 5.6 -8 . 9 1 6.8 0 . 83
( 42 runs )
EHTD 5.5 4. 1 -5 . 0 14.1 0 . 86
Ph am 1 984c
MC P 5.8 4.1 -4.2 1 4. 4 0 . 86
Pham 1 9 8� EHTD 3 .5 5.0 -7 . 3 1 2.9 0 .84
( Eq . ( 7 . ) )
MC P 3.8 5.0 -6 . 6 13.3 0 . 84

FEM'' 1 .5 5.5 -7 . 1 1 1 .3 1 . 00
EIITD 5. 1 5.2 -6 . 7 19.0 0 . 79
FDM
MCP 5.0 5.3 -6. 7 1 9 .2 0 . 80
Two-
D imens ional EHTD 5.2 5.2 -6 . 7 1 9 .2 0 . 85
Irregular Eq. ( 7 . 2 )
MC P 5.2 5.3 -6 . 7 1 9 .3 0 . 86
Tha w i ng
EHTD 5.6 5.4 -7 . 4 16.5 0 . 82
( Tabl e 6 . 6 ) Eq . ( 7 . 3 )
MC P 5.6 5.5 -7 . 4 1 6. 7 0 . 83
( 4 1 runs )
EHTD 6.4 5.6 -7 . 8 1 8.9 0 . 82
Eq. ( 7 . 4 )
MC P 6.3 5.7 -7.8 19.1 0 . 83
EIITD 6.0 5.6 -5. 1 20 . 2 0 . 86
Eq. ( 7 .5 )
MC P 5.9 5.7 -5 . 1 20 . 3 0 . 87

FEM " -10.3 6.7 -1 8 . 2 -2 . 0 1 . 00


EHTD 0 .8 10.2 -1 2 . 8 13.2 0 . 89
FDM
MC P -3 . 9 9.7 -1 6.8 7.9 0 . 89
Eq . ( 7 . 7 ) EIITD 2 .5 8.0 -9 . 0 1 2.0 0 .91
or Eq . ( 7 . 2 )
Pyramid MC P -2 . 3 7.6 -1 3 .2 6.8 0 . 91
( Tabl e 6 . 7 ) Cleland & EIITD 2 .8 7.4 -6 .5 1 1 .1 0 . 95
Ear le 1 98 2 b
( 6 runs ) o r Eq . ( 7 . 3 ) MCP -2 . 0 7 .0 -10.8 5.9 0 . 95
Pham 1 98 4 c EIITD 2.3 8.1 -8. 3 13.2 0 . 93
or Eq. ( 7 . 4 )
MC P -2 . 5 7.6 -1 2 . 5 7.8 0 . 93
Pham 1 98 � EHTD 2.6 9.3 1 0 1 5.9 0 . 95
( Eq . ( 1 ) )
- .

or Eq. ( 7 . 5 ) MC P -2 . 2 8.8 -1 1 .3 1 0 .5 0 . 95

• • • cont inued
Tab le 1 0 . 3 con t i nued • • • 21 3

Data Slab Pred ict ion Geomet r i c 1 Mean Std Dev 2 Min 2 Max 2 Corr 3

Method Fac tor (%) (%) ( %) (%) FDM

FEM'' ( 3 -Da l Gr i d ) -5 . 5 6.8 -1 2. 1 15.3 1 . 00


FEM ,. ( 1 -Da l Gr i d ) 0.4 6.7 -7 .7 9.6 0.91
EIITD 2.3 8.8 -1 3 . 1 13 .6 0 . 72
FDM
MC P 2. 1 8.8 -1 3 . 3 13.4 0.71
Eq. (7 .7 ) EIITD 5.0 5.2 -1 .6 1 3 .7 0 . 92
or Eq . ( 7 . 2 )
Sphere MC P 4 .8 5.2 -1 .8 1 3.5 0 . 92
( Tab le 6 . 7 ) Cleland & EIITD 6.9 7.6 -1 . 2 1 6 .5 0 . 94
Ear le 1 982b
( 6 runs ) or Eq . ( 7 . 3 ) MC P 6.7 7.5 -1 . 4 16.2 0 . 94
Pham 1 98 4 c EIITD 2. 1 7-3 -6 . 9 1 4 .5 0 . 94
or Eq . ( 7 . 4 )
MC P 2.0 7.3 -7 . 1 1 4 .2 0 . 94
Pham 1 9 8� EHTD 3.6 7.4 -6 . 5 1 1 .3 0 . 77
( Eq . ( 7 . ) )
or Eq . ( 7 . 5 ) MC P 3 .5 7.4 -6 . 7 1 1 .1 o .77

FEM ,. 5.2 13.o -1 2 . 9 25 . 1 1 . 00


EHTD 14.4 14.3 -1 1 . 5 28 . 8 0 . 89
FDM
MC P 13.3 1 4. 4 -1 2 . 8 27 . 9 0 . 89
Eq. ( 7 .7 ) EHTD 17.2 1 2.7 -4 . 1 31 .6 0 . 96
or E q . ( 7 . 2 )
Egg MC P 1 6.0 12.8 -5 . 4 30 . 6 0 . 96
( Tabl e 6 . 7 ) Cleland & EHTD 20 . 5 1 3 .5 3.0 39 . 8 0 . 97
Ear le 1 982b
( 6 runs ) or Eq . ( 7 . 3 ) MC P 19.3 1 3.5 1 .6 38 . 8 0 . 97
Pham 1 9 84c EHTD 1 3 .7 1 2. 4 -6 . 9 30 . 2 0 . 98
or Eq . ( 7 . 4 )
MC P 12.5 1 2 .5 -8.2 29 . 3 0 . 98
Pham 1 98 � EIITD 1 7 .0 1 3.5 -4 . 1 34.6 0 . 95
( Eq . ( 7 . ) )
or Eq . ( 7 . 5 ) MCP 15.8 13.6 -5 . 4 33 . 7 0 . 94

• • • cont i nued
Tab l e 1 0 . 3 cont inued • . • 214

Data Slab Pred ict ion Geometr ic 1 Mean Std Dev 2 Min 2 Max 2 Corr 3
Method Factor (%) ( %) (%) (%) FDM

FEM .. 35 . 3 8.9 20 . 8 44 . 2 1 . 00
EHTD 15.3 7.3 6.2 24 .6 0 .88
FDM
MC P 12.3 7.0 2.5 20 . 4 0 . 89
Eq . ( 7 . 7 ) EHTD 1 7 .9 7. 1 7.7 28 . 5 0 . 72
or Eq . ( 7 . 2 )
F i sh MC P 1 4 .8 6.3 5.9 24 . 1 0 .79
( Tabl e 6 . 7 ) Cleland & EHTD 19. 1 10.3 2.9 34 . 8 0 . 67
Ear le 1 98 2 b
( 6 runs ) or Eq . ( 7 . 3 ) MCP 1 6 .0 9.3 1.1 30 . 2 0 73

Pham 1 98 4 c EHTD 1 5.3 8.1 2.2 25 . 1 0 . 74


or Eq . ( 7 . 4 )
MC P 12.3 7.3 0.5 20 . 8 0 . 80
Pham 1 9 83 EHTD 18.1 7.4 4.8 25 . 3 0 . 96
( Eq . ( 7 . 6 ) )
or Eq . ( 7 . 5 ) MC P 1 5 .0 7.0 3.0 21 . 0 0 . 99

1 EHTD from Eq . ( 1 0 . 1 ) , MC P from Eq . ( 1 0 . 2 ) .


2 Std Dev standard dev iat ion , Min
= min imum , Max
= = max imum.
3 Correlat ion coe ff ic ient ( r ) compared wi th the percentage d i fferenc es
for e i ther the full f in i te d i fference me thod or the ful l f in i te
e lement method •

.. Full f i n it e d i fference method or full f i n it e element method .


21 5

11 T ESTI NG OF PRED ICTION MET HODS FOR OTHER


MATER IALS AND DATA SETS

11.1 INTRODUC TION

In Sec . 6 . 1 it was dec ided to exclude var iat i o n o f N ( the ratio o f the
2
unfroz en to the fr ozen volumetric spec ific heat capac i ty ) , and N 3 ( the
ratio of the unfrozen to th e froz en thermal conductiv ity ) , in the ma in
thaw ing e x perimental des ign as the effect o f these factors was expected
to be smal l . Howe ver some t es t ing was requ ired to ensure that th i s was
the case . Also to comp lete the testing of freez ing and thaw i ng t i me
p r ed i ct ion me thods , a l l avai labl e accurate e xper imenta l data - for as
many geome tr i es , cond i t io ns and phase change materials as poss ib le
should be used .

In the l i terature no exper i mental thaw ing data for foods were found for
wh ich the cond i t ions and geome tr ies were suf f ic ien t l y acc urate l y
measured , descr i b ed and cont rol led for them t o b e use d t o test
pr ed ic t ion me thod s . Howe v er , in the present stu dy f our teen thaw i ng
exper iments were conducted w i th minced l e an beef ( Table 6 . 8 ) .

For freez i ng a compreh ens i ve se t o f data for a range o f r egul ar


geome t r i es , food mater ials and f inal thermodynamic centre t emperatur es
ex i s ts ( Cle l and & Ear l e 1 984a , Pham 1 98 4 c ) . The only addi t iona l da ta
collected wer e four freez ing runs using t wo-d imens ional irregular
shape s of m i nc ed lean bee f . The compo sit e fr eez i ng and thawing data
set i s descr ib ed i n Table 1 1 . 1 .

1 1 .2 NUMER IC AL PRED ICT ION METHODS

It was expec ted that the f i n i te d ifference and f i n i t e element nume r i c al


methods wou l d introduce neglig ib l e pred i c t io n inaccuracy when app l i ed
to foods i f the space and t i me step intervals used were suff ic iently
small . Howe ve r , an increase in thermal p roperty data unc er ta i nty
relat ive to Tylose data would be expected ( Sec . 5 . 2 ) . Also the nature
of food materials means tha t experimental error is l ike l y to exc eed
that for exp er i ments using Tylose , wh ich is a homogenous gel
T es t i ng of Pred i c t ion Me thods 216

( Sec . 5 .7 . 3 )

The nume r i c al me thods to pre d ict freez ing and thawi ng t i me s had
prev i ously been compared wi th mo s t of the
compos i t e data set for
freez ing ( Cl e land & Earle 1 98 4a ) and for thaw ing of Ty lose ( Sec . 7 . 2
and Chap . 8 ) . Numer ica l me thod pred i c t ions were mad e for the rema i n i ng
thaw ing dat a for mi nced lean beef a nd the more r ecently publ ished
freez i ng dat a , ( marked in Tab l e 1 1 . 1 ) , to comp l e te the compar i son . The
f i n it e d i fference me thod was used for regular shapes and the f in i te
elemen t me thod for irregul ar shapes . The th erma l p roper ty dat a u sed
wer e those given by Cl e land & Earle ( 1 98 4 a ) . For Ty lose and m i nced
lean bee f these are cons is ten t wi th , a l though s l i ghtly d i f fer e nt to ,
the da ta in Table 5 . 1 . The pred ict ions are summa r i se d i n Table 1 1 . 2
( for thaw i ng o f minced lean bee f ) and Table 1 1 . 4 ( for freez i ng o f
mul t i -d imens ional shap es ) .

Of f se t o f the mean per cen tage d i f ferenc e between predicted and


exper imenta l t i mes from zero was s i g n i f icant for the two-d imens ional
irregu l ar shapes , for f in i t e cyl inder experi me nts carr i ed ou t by de
M i chel is & Calv e lo ( 1 983 ) , and for a ll the data collecte d by Hayakawa
et a l ( 1 9 83 a ) and Succar & Hawakawa ( 1 98 4 ) . Cons istent off set can
ar i se from sy stemat i c exper imental error ( partic ularly in est i mates o f
the surfac e heat transfer c oe f f i c i e nt ) or from thermal p roper ty data
used in calc ulat ions not truly r epresenting the samples o f the test
mater i a l s u sed .

For the two-d imens ional irregular shapes the former was cons ide red most
l ik e l y . Measur ed rates o f temperature change w i th t ime were
con s i s tently under-pred icted by the f in i te e lement method thr oughou t
bo th the freez i ng or thaw i ng proc esses for b oth the shapes u sed
( Fi g . 6 . 1 9 ) . Al though thermocouple placement error may have
contr ibuted to oc ca s ional und er-pred i c t io n , the under-pred i c t io n was
too cons is tent for th is factor alone to expla i n the observed
t emperature prof i l es . Hen ce the assump t i on made in Sec . 5 . 8 . 3 that
both the contac t res i s ta nc e between the Tylose and the mou ld wall and
the contac t res is tan ce b etween the minc ed lean beef and the mould wall
were equal , may not be correct . The d if ference may have ari sen b ec ause
Tylo se i s a dry gel , whereas minced lean beef has a f ib r ou s , mo i s t
Test i ng of Pred ict ion Methods 21 7

nature ( Sec . 5 . 7 . 3 ) .

The o ther data set s for wh ich o ff se t s from z ero were s i g n i f icant were
al l t aken from th e l i terature . It was d if f icu lt t o as sess wh ether
inappro p r i ate therma l dat a , or ex per imental inaccuracy cause d the
problems in these data sets . I n several cases surfac e heat transf er
coe f f ic ients were not determi ned i ndependently of the freez ing
exper i men t s . It is v ery unl ike ly that p roblems in th e f in i te
d ifferenc e or f i n i te element method c alcu lat ions were th e cause as the
pr ograms were t es ted r i gorously ( Chap . 4 ) and i t has been demonstrated
in th is study that they performed well for 90% of the whole data se t .

The standa rd dev i a t ions o f pred i c t ion error s wer e general ly l arger for
food ex per i ments than for Ty lose exper iments ( for example - Ty los e slab
tha w i ng , 3 . 5 % ; min ced lean bee f s l a b th aw i ng , 5 . 9% ) . Th i s r e f l ec ts
the d i f f i c ult ies d isc ussed previously in exper imenting w i th food
mater ials ( Sec . 5 . 2 ) .

It shou ld be noted that some indiv idual se ts o f data are very sma l l
( l ess than s i x runs ) , s o pred i c t io n error means and standard dev i at i ons
calc ulated for these se ts are less mean ing ful than thos e for larg e data
set s . Simi lar l y , when simpl e pred i c t i on methods are eval ua ted in
Sec . 1 1 . 3 , the c orre lat i on coe f f ic ients bet ween these methods and th e
nume r ical method pred ic t i on resu l ts for such s ma l l data se ts shou l d be
v i ewed in the same way .

1 1 .3 SIMPL E PRED ICT ION METHODS

If the e ffec ts of N 2 and N 3 are un imp ortant , results calc ulated us ing
prev i ously-recommended s i mp l e pred i c t ion formulae ( Sec . 7 . 3 . 3 ) wou l d b e
expected to match results for the nume rical methods ( s imi lar mean
pred i c t io n error and h i gh corre lat io n coeff ic ients ) , as wel l a s be i ng
i n general a greemen t w i th the exper i mental t i me s . Very good agreement
wi th al l o f the exper i men t s w ould not b e ex pec ted in the l ig h t of the
poor n ume r i c a l method pred i c t i o ns for some data .

Tab l e 1 1 . 2 summar i ses the pred i c t io n acc uracy for the four simp l e slab
thawi ng t i me predi c t i o n formulae ( Eqs . ( 7 . 2 ) to ( 7 .5 ) ) u se d in
Tes t i ng of Pred i c t ion Meth ods 218

conj unc t io n w ith the two geometr ic factors ( Eqs . ( 1 0 . 1 ) and ( 1 0 . 2 ) ) for
the minced lean beef data ( Tab le 6 .8 ) . The thermal pr oper ty data u sed
ar e g i v en in Table 5 . 2 . As both the mean and standard dev iatio ns o f
the pred i c t ion error s ar e s i mi lar t o tho se for the numer i ca l me thods it
was conc luded that any inacc uracy a r i s ing from n eglect o f N2 and N 3 was
suff ic i en t l y smal l to be ignored .

Cleland & Earle ( 1 984a , 1 98 4 b ) and Pham ( 1 983 , 1 98 4 c ) use d the data of
Cle land & Ear l e ( 1 97 7 a , 1 97 9 a , 1 979b ) , de M i chel i s &
Ca lve lo ( 1 983 ) and
Hung & Thompson ( 1 983 ) to test methods to p redict freez ing t i mes . I t
was cons idered worthwhi le to ca lculate resu l t s equ iva l e nt to Tab l e 1 o f
Cl e land & Earle ( 1 98 4b) for Eq. ( 7 . 7 ) , th e o n e new freez ing me thod
under cons ide rat ion her e . The resu l t s a r e g i ven i n Table 1 1 . 3 . For
data to a f i nal thermodynamic centre temper ature other than - 1 0 °C , the
mod i f icat i on o f Cle land & Ear le ( 1 98 4 b) was used to correc t the
pred iction o f Eq. ( 7 . 7 ) wh ich i s based on a f inal th ermodynamic c entre
temperature o f - 1 0 °C . The therma l property data g iven by Cl e land &
Ea r l e ( 1 984a) were u se d to be cons is ten t w i th the prev i ously publ i sh ed
results .

S i rice the previous stud ies , some new slab data have been pub l i shed by
Succar & Hayakawa ( 1 9 84 ) . The se are a lso i nc l uded in Table 1 1 . 3 . W i th
the thermal property data use d , ne i ther these n i n e runs , nor resul t s
from the same l aboratory for o th er shapes ( Hawak awa e t a l 1 983a ) shown
in Table 1 1 . 4 co uld be predic ted accurately by any of the methods
cons idered , inc lud i ng the f in i te d i f ferenc e me thod .

Compar ison o f Table 1 1 . 3 w i th Table 1 of Cleland & Earle ( 1 98 4b ) shows


that E q . ( 7 . 7 ) was equa l l y good as the best pr ev i ously pr opo sed me thods
o f Cle land & Earle ( 1 984b) , and Pham ( 1 983 , 1 98 4 a , 1 98 4c ) ( Sec . 7 . 3 . 3 ) .
Ne i ther ignor ing the effects o f N 2 and N 3 nor us ing the me thod o f
Cleland & Earl e ( 1 98 4 b ) to acc ount for d if fere nt f ina l temperatures
s i g n i f icantly a ffected the pred ic tion acc ur acy .

The four s i mp l e slab freez ing and thaw i ng t i me pred i c t i on approaches


d i sc ussed in th is sec t i o n and in Sec . 7 . 3 , c an all be extended to any
o th er shape b y e i the r o f the geometric factors EHTD or MCP . Both
geometric fac tors have been tested for most of the T y lo s e
Tes t i ng of Pred ict ion Methods 21 9

mu l t i -d imens io na l shap e freez ing and tha w i ng data ( Chap . 1 0 and Tab l e
1 1 .2) . In Table 1 1 . 4 , the pred ict ion results of these s imple
pred i c t ion me thods ar e summar i se d for the rema in i ng mu l t i -d imens iona l
shap e dat a from the compos i te data se t of Tab le 1 1 . 1 . Becau se a l l four
s l ab approaches are app rox imate ly equ ival en t , only one was used in th is
analy si s . The method of Pham ( 1 98 4 c ) was arb itra r i ly chos en . Aga in ,
th e s imp l e pred ict i on me thod resu l ts c l ose ly fol low the numer i ca l
method pred ictions and ag ree a s well as cou ld b e expec ted w i th the
ex per imental dat a . No s ign i ficant d i f ferenc e in pred ict ion acc uracy
between MC P and EHTD cou ld be d iscerned .

1 1 .4 SUMMARY

Numer ical freez ing and thaw ing t ime pred ict ion me thods were sl igh t ly
l ess accurat e for food mater ials than for Tylo se due to l arger thermal
property dat a and expe r i mental uncertaint i es .

S i mp l e slab pred i c t i on me thods and geometric factor s deve loped to


pred ic t fr eez ing and thawi ng t i mes for Ty lose ·have b een shown to be
equa l l y app l ic able to food mate r i a ls and for a wid er range of
cond i tions and shapes than previously cons idered .

The effects o f N 2 and N 3 on freez ing t i me s were suff i c iently sma l l


compared wi th e xpe r i men tal uncerta inty , that n o j u s t i f icat i on ex i s ted
for inclu s io n o f terms i nvolv ing N 2 and N 3 in the s i mple emp ir ica l
me thods . Fewer test da ta ex i st for tha w i ng , but there is no ind ic a t i on
from these that N 2 and N 3 ar e any more s i gn i f icant for thawi ng than for
freez i ng .
Tes t i ng of Pred ic t ion Methods 220

Tab l e 1 1 1 Composi te Data Set For Tes t ing of Freez i ng and Tha w i ng Ti me

Pr ed ict ion Me thods

Type So urc e Mater i a l Shape 1 Fina l Centre No . of


Temperature Runs
( oc )

Tab le 6 . 2 Tylose Slab 0.0 35


Tab l e 6 . 3 Ty l ose Cy l ind er 0 .0 34
Table 6 . 4 Ty lose Sphere 0 .0 35
Table 6 . 5 Ty lo se Br i ck 0.0 68
Thaw ing Table 6 . 6 Ty lose 2D I 0 .0 41
Tab l e 6 . 7 Ty l ose 3D I 0 .0 11
*
Tab l e 6 . 8 Lean Bee f Slab 0 .0 6
*
Table 6 . 8 Lean Beef Br i ck 0 .0 4
Table 6 . 8 Lean Bee f 2DI 0 .0 4*

Tab l e 6 . 6 Tylose 2DI -10.0 42


Fr eez ing Table 6 . 7 Ty lose 3D I -1 0 .0 13
*
Table 6 . 8 Lean Beef 2DI - 1 0 .0 4

Tylose Slab -1 0 . 0 43
Cleland & Ear l e 1 977a Le an Bee f Slab - 1 0 .0 6
Po tato Slab -1 0 . 0 6
Freez i ng
Cl e land & Ear le 1 97 9 a Ty lose Cy l inder - 1 0 .0 30
Cleland & Ear l e 1 979a Tylo se Sph ere -10.0 30
Cl e land & Ear le 1 97 9b Ty lose Br ick - 1 0 .0 72

Tylos e Slab -1 8 .0 23
Lean B ee f Slab - 1 8 .0 9
Freez ing Hung & Thompson 1 983 Potato Slab - 1 8 .0 9
Ground Beef Slab -18.0 9
Car p Slab -1 8 .0 9

• • • continued
Tab l e 1 1 . 1 con t inued • • • 221

Ty pe Sourc e Ma ter ial Shape 1 F i na l Centre No . of


Temperature Runs
( oc )

Lean Beef Slab - 1 8 .0 5


de Miche l i s & Lean Beef F in i te Cy l . -18.0 4
Freez ing Ca lve lo 1 983
Lean Beef In f . Rod - 1 8 .0 3
Lean Beef Br i ck -10.0 17
*
Lean Bee f In f . Rod - 1 5 .0 6
Freez i ng Hayakawa et a l 1 983a
Tylo se F in i te Cy l . - 1 5 .0 6*
*
Freez i ng Succar & Ty lose Slab Var iou s 9
Hayakawa 1 9 84

THAW ING Total 238


FR EE ZING Tota l 355

FR EEZI NG AND TH AW ING TOTAL 593

1 2DI two-d imens ional irregu lar shapes , 3D I


= three-d imens io nal
irr egular shapes .
*
Data not prev iously used to test pred ic t i o n methods .
Tes t ing of Pred i c t ion Methods 222

Tab le 1 1 . 2 Summary of Perc entage D i f ferences Between Expe r i mental


Thaw ing T i mes and Pred ic ted Thaw ing Time s For Minced Lean
Beef

Data Slab Pred iction Geometr ic 1 Mean Std De v 2 Min 2 Max 2 Corr 3
Method Factor ( %) ( %) ( %) (%) FDM

FDM " - 4 .5 5.9 -1 1 9


• 2.5 1 . 00
Slab Eq. (7 . 2 ) -7 . 2 8 .5 -1 5 . 4 3 .6 0 .81
( Tab le 6 . 8 ) Eq . ( 7 . 3 ) -7 .0 6.9 - 1 6 .0 1 1
• 0.91
( 6 runs ) Eq. (7 . 3 ) -6 . 4 7 .9 -1 9 . 4 3 .6 0 . 92
Eq. ( 7 . 5 ) -5 . 4 6 .0 -1 0 .6 2 .5 0 . 96

FDM " 4 .8 7 .0 -3.0 1 7 .3 1 .00


EHTD 1 .8 6.0 - 3 .5 10.1 1 . 00
Eq. (7 .2)
MC P 1 .8 6 .0 -4.1 9.4 0 . 99
Rect angu l ar
Br ick EHTD 1 .7 8 .5 -5 . 3 14 1
• 0 .99
Eq. ( 7 . 3 )
( Table 6 . 8 ) MC P 1 .7 8.3 -5.9 1 3.4 0 . 99
( 4 runs ) EHTD -0 . 6 5 .5 -6 . 1 6 .3 0 . 97
Eq. ( 7 . 4 )
MC P -0 . 7 5 .5 -7 . 5 5.6 0 . 93
EHTD 2 .5 4 .6 -2 . 1 8 .6 0 . 94
E q . ( 7 .5 )
MC P 2.4 4.3 -1 .5 7 .9 0 . 95

F EM'' 1 3 .8 3.0 9 .5 16.2 1 . 00


EHTD 1 4 .6 2.8 1 2. 4 1 8 .8 0.16
Eq. (7 . 2 )
Two- MC P 1 4 .8 2 .8 1 2 .5 1 8 .9 0.1 6
D i mens iona l
Irregular EHTD 1 5 .7 3 .7 1 1 .6 1 9 .9 0 . 55
Eq. ( 7 . 3 )
( Tab le 6 . 8 ) MCP 1 5 .8 3 .7 1 1 .8 20 . 0 0 .56
( 4 run s ) EHTD 1 7 .0 3 .3 1 3 .5 21 . 0 0 . 76
Eq . ( 7 . 4 )
MC P 17.2 3 .3 1 3 .6 21 .2 0 . 76
EHTD 1 6 .9 1 .4 15.4 1 8 .5 0 . 81
Eq. (7 .5 )
MC P 1 7 .0 1 .4 15.6 1 8 .7 0 .82

1 EHTD calc ulated from E q . ( 1 0 . 1 ) , MCP ca lculated from Eq. ( 1 0 . 2 ) .


2 Std De v standard dev iation, Min min i mum, Max : max i mum.
• •

' Correlat ion coeff ic ient ( r ) compared w ith the perc entage d if ferences
for ei ther the full f in i te d i fferen ce method or the ful l f ini te
element method .
" Full f in i t e d iffer ence method or full f in it e f i ni t e e lement method .
Tes t ing of Pred i c t ion Methods 223

Table 1 1 . 3 Summary o f Perc entage D i f ferences Between Exper i me ntal


Fr eez ing T i me s and Predicted Freez ing Times Calcu lated By
Eq. ( 7 . 7 )

Sou rc e Shape Mate r i a l Geometric Me an Std Dev 2 Min 2 Max 2 Corr 3


1
Factor No . ( % ) (% ) (%) (%) FDM

Slab Ty lose 43 0.8 2 .8 -5. 9 5.7 0 . 32


Cle l and & Slab Lean Beef 6 2.9 4 .0 -1 .8 9 .4 0 . 64
Ear le 1 97 7 a
Slab Pota to 6 -0 . 5 2.2 -3 . 2 2 .2 0 .35
Cleland & Cyl inder Tylo se 30 -1 . 4 4.0 -9 . 8 5.3 0 .81
Ear le 1 979a
Sphere Ty lose 30 0.5 4 .0 -7 . 2 1 0 .6 0 . 92
Cle l and & EHTD 72 0 .3 5. 1 -8 5 1 0 . 1
• 0 . 56
Ear le 1 97 9 b Br ick Tylose
MC P 72 -1 3
• 5.4 -1 1 . 1 8.3 0 . 65

Ty lose 23 -3 . 9 8 .0 -24 .8 9 .2 0 .78


Lean Beef 9 - 4 .8 9 .0 -17.2 9 .3 0 . 89
Hung &
Thomp so n Slab Po tato 9 - 8 .0 4 .9 - 1 6 .5 0.0 0 . 20
1 983
Ground Bee f 9 2 .8 1 5 .8 - 2 1 . 6 22 . 2 0 . 93
Carp 9 -1 7
• 1 2.1 -22 . 2 1 6 . 7 0 . 89

Slab Lean Beef 5 4.4 4 .6 -1 5 1 0 . 9


• 0 .91
de M i che l is
& Calvelo F in i te EHTD 24 8 .7 1 1 .5 - 1 2 . 0 29 . 3 0 . 84
1 98 3 Cyl inder Le an Beef
Rod & MC P 24 7.4 1 0 .8 - 1 2 . 3 29 .0 0 .87
Br i ck

Succar & Slab Tylose 9 3 1 .5 8.3 2 3 . 0 43 . 1 1 . 00


Ha yakawa
1 9 84

1 EHTD c alc ulated from Eq. ( 1 0 . 1 ) , MC P calc ulated from Eq. ( 1 0 . 2 ) .


2 Std Dev standard deviat ion, Min
s m i n i mum, Max
= max imum. =

3 Correlat ion coeff ic ient ( r ) compar ed w ith th e perc entage d if ferences


for the f ini te d if ferenc e method .
Test ing o f Pr ed ict ion Methods 22 4

Tab l e 1 1 . 4 Summary of Perc ent age D i f ferences Between Exper i me ntal


Freez i ng Ti mes For Mu lti-Di mensional Shapes and Freez i ng
T i mes Ca lcu lated By S i mple Prediction Methods 1

Sou rce Shape Mate r i a l Geometr i c Mean Std Dev 3 M i n 3 Max 3 Corr �
Fac tor 2 (%) (%) (% ) ( % ) FOM

I n fi n i te FDM 5 -2 . 7 3 .9 -6.7 1.0 1 .oo


Rod
Lean Beef EHTD -5 . 3 9.8 -1 6.6 0 .5 0 .91
(3 runs )
MC P -5 . 6 9 .7 -1 6.8 0.2 0.91
F i n ite FDM 5 1 3 .6 9.3 6.1 25 . 6 1 .0 0
d e M i che l i s Cy l inder
& Ca lvelo Lean Beef EHTD 1 4.4 8.3 10.6 30 . 0 0 . 97
1 983a ( 4 runs )
MC P 18.2 8.9 10. 1 29 7 1 00
• •

Rec tangu lar FDM 5 8 .6 6 .5 -4 . 2 1 7 .7


1 00

Br ick
Lean Beef EHTD 4.2 6.9 - 1 2 .6 1 3 .6 0 .84
( 1 7 runs )
MC P 2.8 6 .0 - 1 2 .5 1 0 .0 0 .89

Infin i te FDM 5 23.8 8 .0 1 4.9 37 . 9 1 .00


Rod
Lean Beef EHTD 28 . 8 8.2 1 9 .5 43 . 0 1 .00
( 6 runs )
Hay akawa MC P 28 . 4 8.3 18 9
• 42 . 7 1 00•

et a l
1 98 3 Finite FDM 5 66 . 9 1 9 .9 45 . 2 9 1 . 5 1 .00
Cyl inder
Ty lose EHTD 80 . 8 21 . 5 57 . 5 1 07 . 3 1 .0 0
( 6 run s )
MC P 80 . 9 22 . 3 56. 8 1 08 2 • 1 00•

1 0.8 2 .2 8.2 1 3 .5 1 .00


20I 6
Tabl e 6 . 8 Lean Beef EHTD 12.5 1 .8 10.5 1 4 .6 0 . 46
( 4 runs )
MC P 1 3. 2 2 .3 1 0 .7 1 5 .9 0 . 37

1 Us i ng ?ham ' s ( 1 9 8 4 c ) slab predi c t io n method for fr eez ing .


2 EHTD c alculated from Eq. ( 1 0 . 1 ) , MC P ca lculated from Eq. ( 1 0 . 2 ) .
3 Std Dev • s ta ndard devia t i o n , Min m i n i mum, Max
= max imum. =

� Corre la t i o n coeffic ient ( r ) compared w i th the percentage d if ferences


for e i ther the ful l f ini te d if ferenc e me thod or the fu l l f in i te
element method .
5 Ful l f in i te d ifference method or ful l f in i t e e lement method .
6 20 I • t wo-d i mens i o nal i r r e gular shapes .
225

12 OVERALL EVALU AT ION OF FREEZING AND THAWING


TIME PREDICTION METH